Sylvain Defresne reports that libexif is vulnerable to a buffer overflow vulnerability due to insufficient input checking. This could lead crash of applications using libexif.

Neo Security Team reports:

If we specify a variable in the html code (any type: hidden, text, radio, check, etc) with the name allowhtml, allowbbcode or allowsmilies, is going to be on the html, bbcode and smilies in our signature.

This is a low risk vulnerability that allows users to bypass forum-wide configuration.

Two separate SQL injection vulnerabilites have been identified in the PostNuke PHP content management system. An attacker can use this vulnerability to potentially insert executable PHP code into the content management system (to view all files within the PHP scope, for instance). Various other SQL injection vulnerabilities exist, which give attackers the ability to run SQL queries on any tables within the database.

A cross-site scripting vulnerability is present in the PostNuke PHP content management system. By passing data injected through exploitable errors in input validation, an attacker can insert code which will run on the machine of anybody viewing the page. It is feasible that this attack could be used to retrieve session information from cookies, thereby allowing the attacker to gain administrative access to the CMS.

Two exploits have been identified in the Linux RealPlayer client. RealNetworks states:

RealNetworks, Inc. has addressed recently discovered security vulnerabilities that offered the potential for an attacker to run arbitrary or malicious code on a customer's machine. RealNetworks has received no reports of machines compromised as a result of the now-remedied vulnerabilities. RealNetworks takes all security vulnerabilities very seriously.

The specific exploits were:

• Exploit 1: To fashion a malicious WAV file to cause a buffer overflow which could have allowed an attacker to execute arbitrary code on a customer's machine.
• Exploit 2: To fashion a malicious SMIL file to cause a buffer overflow which could have allowed an attacker to execute arbitrary code on a customer's machine.

Tavis Ormandy reports:

magemagick-6.2.0-3 fixes an potential issue handling malformed filenames, the flaw may affect webapps or scripts that use the imagemagick utilities for image processing, or applications linked with libMagick.

This vulnerability could crash ImageMagick or potentially lead to the execution of arbitrary code with the permissions of the user running ImageMagick.

The uim developers reports:

Takumi ASAKI discovered that uim always trusts environment variables. But this is not correct behavior, sometimes environment variables shouldn't be trusted. This bug causes privilege escalation when libuim is linked against setuid/setgid application. Since GTK+ prohibits setuid/setgid applications, the bug appears only in 'immodule for Qt' enabled Qt. (Normal Qt is also safe.)

The lighttpd website reports:

In lighttpd 1.3.7 and below it is possible to fetch the source files which should be handled by CGI or FastCGI applications.

The vulnerability is in the handling of urlencoded trailing NUL bytes. Installations that do not use CGI or FastCGI are not affected.

The phpbb developer group reports:

phpBB Group announces the release of phpBB 2.0.13, the "Beware of the furries" edition. This release addresses two recent security exploits, one of them critical. They were reported a few days after .12 was released and no one is more annoyed than us, having to release a new version ini such a short period of time. Fortunately both fixes are easy and in each case just one line needs to be edited.

Two iDEFENSE Security Advisories reports:

An exploitable stack-based buffer overflow condition exists when using NT Lan Manager (NTLM) authentication. The problem specifically exists within Curl_input_ntlm() defined in lib/http_ntlm.c.

Successful exploitation allows remote attackers to execute arbitrary code under the privileges of the target user. Exploitation requires that an attacker either coerce or force a target to connect to a malicious server using NTLM authentication.

An exploitable stack-based buffer overflow condition exists when using Kerberos authentication. The problem specifically exists within the functions Curl_krb_kauth() and krb4_auth() defined in lib/krb4.c.

Successful exploitation allows remote attackers to execute arbitrary code under the privileges of the target user. Exploitation requires that an attacker either coerce or force a target to connect to a malicious server using Kerberos authentication.

The Cyrus IMAP Server ChangeLog states:

• Fix possible single byte overflow in mailbox handling code.
• Fix possible single byte overflows in the imapd annotate extension.
• Fix stack buffer overflows in fetchnews (exploitable by peer news server), backend (exploitable by admin), and in imapd (exploitable by users though only on platforms where a filename may be larger than a mailbox name).

The 2.1.X series are reportedly only affected by the second issue.

These issues may lead to execution of arbitrary code with the permissions of the user running the Cyrus IMAP Server.

Debian Security Advisory reports:

jaguar@felinemenace.org discovered a format string vulnerability in sup, a set of programs to synchronize collections of files across a number of machines, whereby a remote attacker could potentially cause arbitrary code to be executed with the privileges of the supfilesrv process (this process does not run automatically by default).

A Mozilla Foundation Security Advisory reports:

A predictable name is used for the plugin temporary directory. A malicious local user could symlink this to the victim's home directory and wait for the victim to run Firefox. When Firefox shuts down the victim's directory would be erased.

A Mozilla Foundation Security Advisory reports:

Plugins (such as flash) can be used to load privileged content into a frame. Once loaded various spoofs can be applied to get the user to interact with the privileged content. Michael Krax's "Fireflashing" example demonstrates that an attacker can open about:config in a frame, hide it with an opacity setting, and if the attacker can get the victim to click at a particular spot (design some kind of simple game) you could toggle boolean preferences, some of which would make further attacks easier.

The "firescrolling" example demonstrates arbitrary code execution (in this case downloading a file) by convincing the user to scroll twice.

Workaround: Disable JavaScript.

The version 0.06_1 and prior have a format string vulnerability which can be triggered by using a carefully-crafted BDF font file.

psoTFX reports:

phpBB Group are pleased to announce the release of phpBB 2.0.12 the "Horray for Furrywood" release. This release addresses a number of bugs and a couple of potential exploits. [...] one of the potential exploits addressed in this release could be serious in certain situations and thus we urge all users, as always, to upgrade to this release as soon as possible. Mostly this release is concerned with eliminating disclosures of information which while useful in debug situations may allow third parties to gain information which could be used to do harm via unknown or unfixed exploits in this or other applications.

The ChangeLog for phpBB 2.0.12 states:

• Prevented full path display on critical messages
• Fixed full path disclosure in username handling caused by a PHP 4.3.10 bug - AnthraX101
• Added exclude list to unsetting globals (if register_globals is on) - SpoofedExistence
• Fixed arbitrary file disclosure vulnerability in avatar handling functions - AnthraX101
• Fixed arbitrary file unlink vulnerability in avatar handling functions - AnthraX101
• Fixed path disclosure bug in search.php caused by a PHP 4.3.10 bug (related to AnthraX101's discovery)
• Fixed path disclosure bug in viewtopic.php caused by a PHP 4.3.10 bug - matrix_killer

Ulf Härnhammar reports multiple security vulnerabilities in unace-1.2b:

• There are buffer overflows when extracting, testing or listing specially prepared ACE archives.
• There are directory traversal bugs when extracting ACE archives.
• There are also buffer overflows when dealing with long (>17000 characters) command line arguments.

Simon Tatham reports:

This version fixes a security hole in previous versions of PuTTY, which can allow a malicious SFTP server to attack your client. If you use either PSCP or PSFTP, you should upgrade. Users of the main PuTTY program are not affected. (However, note that the server must have passed host key verification before this attack can be launched, so a man-in-the-middle shouldn't be able to attack you if you're careful.)

Davide Madrisan reports:

The `dcopidlng' script in the KDE library package (kdelibs-3.3.2/dcop/dcopidlng/dcopidlng) creates temporary files in a unsecure manner.

Note: dcopidlng is only used at build time, so only users installing KDE are vulnerable, not users already running KDE.

A Debian Security Advisory reports:

Ulf Härnhammer from the Debian Security Audit Project discovered a format string vulnerability in bidwatcher, a tool for watching and bidding on eBay auctions. This problem can be triggered remotely by a web server of eBay, or someone pretending to be eBay, sending certain data back.

A Debian Security Advisory reports:

Albert Puigsech Galicia discovered a directory traversal vulnerability in a proprietary FTP client (CAN-2004-1376) which is also present in gftp, a GTK+ FTP client. A malicious server could provide a specially crafted filename that could cause arbitrary files to be overwritten or created by the client.

A Secunia Advisory reports:

Michael Holzt has discovered a vulnerability in Opera, which can be exploited by malicious people to trick users into executing malicious files.

The vulnerability is caused due to an error in the processing of "data:" URIs, causing wrong information to be shown in a download dialog. This can be exploited by e.g. a malicious website to trick users into executing a malicious file by supplying a specially crafted "data:" URI.

Giovanni Delvecchio reports:

Opera for linux uses "kfmclient exec" as "Default Application" to handle saved files. This could be used by malicious remote users to execute arbitrary shell commands on a target system.

The PL/PgSQL parser in postgresql is vulnerable to several buffer overflows. These could be exploited by a remote attacker to execute arbitrary code with the permissions of the postgresql server by running a specially crafted query.

Several input validation errors exist in AWStats that allow a remote unauthenticated attacker to execute arbitrary commands with the priviliges of the web server. These programming errors involve CGI parameters including loadplugin, logfile, pluginmode, update, and possibly others.

Additionally, the debug and other CGI parameters may be used to cause AWStats to disclose AWStats and system configuration information.

PowerDNS is vulnerable to a temporary denial-of-service vulnerability that can be triggered using a random stream of bytes.

Max Vozeler discovered several format string vulnerabilities in the movemail utility of Emacs. They can be exploited when connecting to a malicious POP server and can allow an attacker can execute arbitrary code under the privileges of the user running Emacs.

A No System Group security advisory reports that ngircd is vulnerable to a format string vulnerability in the Log_Resolver() function of log.c, if IDENT support is enabled. This could allow a remote attacker to execute arbitrary code with the permissions of the ngircd daemon, which is root by default.

Note: By default the FreeBSD ngircd port does not enable IDENT support.

Florian Westphal discovered a buffer overflow in ngircd which can be used remotely crash the server and possibly execute arbitrary code.

Mark J Cox reports:

Graham Dumpleton discovered a flaw which can affect anyone using the publisher handle of the Apache Software Foundation mod_python. The publisher handle lets you publish objects inside modules to make them callable via URL. The flaw allows a carefully crafted URL to obtain extra information that should not be visible (information leak).

A directory traversal vulnerability in mailman allow remote attackers to read arbitrary files due to inadequate input sanitizing. This could, among other things, lead remote attackers to gaining access to the mailman configuration database (which contains subscriber email addresses and passwords) or to the mail archives for private lists.

Erik Sjölund discovered several issues in enscript: it suffers from several buffer overflows, quotes and shell escape characters are insufficiently sanitized in filenames, and it supported taking input from an arbitrary command pipe, with unwanted side effects.

John Heasman and others disovered that non-privileged users could use the LOAD extension to load arbitrary libraries into the postgres server process space. This could be used by non-privileged local users to execute arbitrary code with the privileges of the postgresql server.

An Ethreal Security Advisories reports:

Issues have been discovered in the following protocol dissectors:

• The COPS dissector could go into an infinite loop. CVE: CAN-2005-0006
• The DLSw dissector could cause an assertion. CVE: CAN-2005-0007
• The DNP dissector could cause memory corruption. CVE: CAN-2005-0008
• The Gnutella dissector could cuase an assertion. CVE: CAN-2005-0009
• The MMSE dissector could free statically-allocated memory. CVE: CAN-2005-0010
• The X11 dissector is vulnerable to a string buffer overflow. CVE: CAN-2005-0084

Impact: It may be possible to make Ethereal crash or run arbitrary code by injecting a purposefully malformed packet onto the wire or by convincing someone to read a malformed packet trace file.

The squid patches page notes:

This patch addresses a HTTP protocol mismatch related to oversized reply headers. In addition it enhances the cache.log reporting on reply header parsing failures to make it easier to track down which sites are malfunctioning.

It is believed that this bug may lead to cache pollution or allow access controls to be bypassed.

According to Python Security Advisory PSF-2005-001,

The Python development team has discovered a flaw in the SimpleXMLRPCServer library module which can give remote attackers access to internals of the registered object or its module or possibly other modules. The flaw only affects Python XML-RPC servers that use the register_instance() method to register an object without a _dispatch() method. Servers using only register_function() are not affected.

On vulnerable XML-RPC servers, a remote attacker may be able to view or modify globals of the module(s) containing the registered instance's class(es), potentially leading to data loss or arbitrary code execution. If the registered object is a module, the danger is particularly serious. For example, if the registered module imports the os module, an attacker could invoke the os.system() function.

Note: This vulnerability affects your system only if you're running SimpleXMLRPCServer-based server. This isn't harmful at all if you don't run any internet server written in Python or your server doesn't serve in XML-RPC protocol.

Kevin Finisterre discovered bugs in perl's I/O debug support:

• The environmental variable PERLIO_DEBUG is honored even by the set-user-ID perl command (usually named sperl or suidperl). As a result, a local attacker may be able to gain elevated privileges. (CAN-2005-0155)
• A buffer overflow may occur in threaded versions of perl when the full pathname of the script being executed is very long. (CAN-2005-0156).

Note: By default, no set-user-ID perl binary is installed. An administrator must enable it manually at build time with the ENABLE_SUIDPERL port flag.

The newsgrab script uses insecure permissions during the creation of the local output directory and downloaded files.

After a file is created, permissions on it are set using the mode value of the newsgroup posting. This can potentially be a problem when the mode is not restrictive enough. In addition, the output directory is created with world writable permissions allowing other users to drop symlinks or other files at that location.

The newsgrab script creates files by using the names provided in the newsgroup messages in a perl open() call. This is done without performing any security checks to prevent a directory traversal. A specially crafted newsgroup message could cause newsgrab to drop an attachment anywhere on the file system using the permissions of the user running the script.

The newspost program uses a function named socket_getline to read server responses from the network socket. Unfortunately this function does not check the length of the buffer in which the read data is stored and only stops reading when a newline character is found.

A malicious NNTP server could use this bug to cause a buffer overflow by sending an overly long response. Such an overflow allows arbitrary code to be executed, with the privileges of the newspost process, on the affected systems.

The newsfetch program uses the sscanf function to read information from server responses into static memory buffers. Unfortunately this is done without any proper bounds checking. As a result long server responses may cause an overflow when a newsgroup listing is requested from an NNTP server.

According to the Squid Proxy Cache Security Update Advisory SQUID-2005:3,

The WCCP recvfrom() call accepts more data than will fit in the allocated buffer. An attacker may send a larger-than-normal WCCP message to Squid and overflow this buffer.

Severity:

The bug is important because it allows remote attackers to crash Squid, causing a disription in service. However, the bug is exploitable only if you have configured Squid to send WCCP messages to, and expect WCCP replies from, a router.

Sites that do not use WCCP are not vulnerable.

Note that while the default configuration of the FreeBSD squid port enables WCCP support in general, the default configuration supplied does not actually configure squid to send and receive WCCP messages.

An iDEFENSE Security Advisory reports:

Remote exploitation of a buffer overflow vulnerability in the xpdf PDF viewer included in multiple Unix and Linux distributions could allow for arbitrary code execution as the user viewing a PDF file.

The vulnerability specifically exists due to insufficient bounds checking while processing a PDF file that provides malicious values in the /Encrypt /Length tag. The offending code can be found in the Decrypt::makeFileKey2 function in the source file xpdf/Decrypt.cc.

Martin Joey Schulze reports:

Erik Sjöund discovered that zhcon, a fast console CJK system using the Linux framebuffer, accesses a user-controlled configuration file with elevated privileges. Thus, it is possible to read arbitrary files.

When installed from the FreeBSD Ports Collection, zhcon is installed set-user-ID root.

Martin Joey Schulze reports:

Max Vozeler discovered an integer overflow in the helper application camel-lock-helper which runs setuid root or setgid mail inside of Evolution, a free groupware suite. A local attacker can cause the setuid root helper to execute arbitrary code with elevated privileges via a malicious POP server.

An LSS Security Advisory reports:

When a denial of service attack is detected, mod_dosevasive will, among other things, create a temporary file which it will use to trace actions from the offensive IP address. This file is insecurely created in /tmp and it's name is easily predictable.

It is then easy for an attacker to create arbitrary files in any directory that the user under which apache runs has privileges to write.

[...] once the target file is opened, there is a race attack (although difficult to exploit) which can lead to mod_dosevasive overwriting any file that the user under which apache runs has privileges to write.

The squid patches page notes:

This patch makes Squid considerably stricter while parsing the HTTP protocol.

1. A Content-length header should only appear once in a valid request or response. Multiple Content-length headers, in conjunction with specially crafted requests, may allow Squid's cache to be poisioned with bad content in certain situations.
2. CR characters is only allowed as part of the CR NL line terminator, not alone. This to ensure that all involved agrees on the structure of HTTP headers.
3. Rejects requests/responses that have whitespace in an HTTP header name.

To enable these strict parsing rules, update to at least squid-2.5.7_9 and specify relaxed_header_parser off in squid.conf.

A Bugzilla advisory states:

This advisory covers a single cross-site scripting issue that has recently been discovered and fixed in the Bugzilla code: If a malicious user links to a Bugzilla site using a specially crafted URL, a script in the error page generated by Bugzilla will display the URL unaltered in the page, allowing scripts embedded in the URL to execute.

A Secunia Research advisory reports:

Secunia Research has reported a vulnerability in multiple browsers, which can be exploited by malicious people to spoof the content of websites.

The problem is that a website can inject content into another site's window if the target name of the window is known. This can e.g. be exploited by a malicious website to spoof the content of a pop-up window opened on a trusted website.

Secunia has constructed a test, which can be used to check if your browser is affected by this issue: http://secunia.com/multiple_browsers_window_injection_vulnerability_test/

A workaround for Mozilla-based browsers is available.

Manigandan Radhakrishnan discovered a security vulnerability in YAMT which can lead to execution of arbitrary commands with the privileges of the user running YAMT when sorting based on MP3 tags. The problem exist in the id3tag_sort() routine which does not properly sanitize the artist tag from the MP3 file before using it as an argument to the mv command.

According to a whitepaper published by Sanctum, Inc., it is possible to mount cache poisoning attacks against, among others, squid proxies by inserting false replies into the HTTP stream.

The squid patches page notes:

This patch additionally strengthens Squid from the HTTP response attack described by Sanctum.

A Hyperdose Security Advisory reports:

Horde contains two XSS attacks that can be exploited through GET requests. Once exploited, these requests could be used to execute any javascript commands in the context of that user, potentially including but not limited to reading and deleting email, and stealing auth tokens.

Andrew V. Samoilov reported several vulnerabilities that were corrected in MidnightCommand 4.6.0:

• Format string issues (CAN-2004-1004)
• Buffer overflows (CAN-2004-1005)
• Denial-of-service, infinite loop (CAN-2004-1009)
• Denial-of-service, corrupted section header (CAN-2004-1090)
• Denial-of-service, null pointer dereference (CAN-2004-1091)
• Freeing unallocated memory (CAN-2004-1092)
• Using already freed memory (CAN-2004-1093)

Jeroen van Wolffelaar reports that the Perl module File::Path contains a race condition wherein traversed directories and files are temporarily made world-readable/writable.

Marc Schoenefeld reports:

Opera 7.54 is vulnerable to leakage of the java sandbox, allowing malicious applets to gain unacceptable privileges. This allows them to be used for information gathering (spying) of local identity information and system configurations as well as causing annoying crash effects.

Opera 754 [sic] which was released Aug 5,2004 is vulnerable to the XSLT processor covert channel attack, which was corrected with JRE 1.4.2_05 [released in July 04], but in disadvantage to the users the opera packaging guys chose to bundle the JRE 1.4.2_04 [...]

Internal pointer DoS exploitation: Opera.jar contains the opera replacement of the java plugin. It therefore handles communication between javascript and the Java VM via the liveconnect protocol. The public class EcmaScriptObject exposes a system memory pointer to the java address space, by constructing a special variant of this type an internal cache table can be polluted by false entries that infer proper function of the JSObject class and in the following proof-of-concept crash the browser.

Exposure of location of local java installation Sniffing the URL classpath allows to retrieve the URLs of the bootstrap class path and therefore the JDK installation directory.

Exposure of local user name to an untrusted applet An attacker could use the sun.security.krb5.Credentials class to retrieve the name of the currently logged in user and parse his home directory from the information which is provided by the thrown java.security.AccessControlException.

A sudo bug report says:

sudo doesn't unset the CDPATH variable, which leads to possible security problems.

An iDEFENSE Security Advisory states:

Multiple vulnerabilities have been found in Fcron.

1. File contents disclosure
2. Configuration Bypass Vulnerability
3. File Removal and Empty File Creation Vulnerability
4. Information Disclosure Vulnerability

An NGSSoftware Insight Security Research Advisory reports:

Two vulnerabilities have been discovered in RealPlayer which may potentially be leveraged to allow remote code execution, or may used in combination with the Real Metadata Package File Deletion vulnerability to reliably delete files from a users system.

Pavel Kankovsky reports:

Imlib affected by a variant of CAN-2004-0782 too.

I've discovered more vulnerabilities in Imlib (1.9.13). In particular, it appears to be affected by a variant of Chris Evans' libXpm flaw #1 (CAN-2004-0782, see http://scary.beasts.org/security/CESA-2004-003.txt). Look at the attached image, it kills ee on my 7.3.

The flaws also affect imlib2.

eGroupWare contains a bug in the JiNN component that allows a remote attacker to download arbitrary files.

An advisory published by Richard Stanway describes numerous critical vulnerabilities in the Quake II engine:

Due to unchecked input at various stages in the server, remote users are able to cause the server to crash, reveal sensitive information or potentially execute arbitrary code.

Konversation comes with Perl scripts that do not properly escape shell characters on executing a script. This makes it possible to attack Konversation with shell script command injection.

The LDAP authentication helper did not strip leading or trailing spaces from the login name. According to the squid patches page:

LDAP is very forgiving about spaces in search filters and this could be abused to log in using several variants of the login name, possibly bypassing explicit access controls or confusing accounting.

Workaround: Block logins with spaces

	    acl login_with_spaces proxy_auth_regex [:space:]
		    http_access deny login_with_spaces
	    

Kenshi Muto discovered that the CUPS server would enter an infinite loop when processing a URL containing /...

A US-CERT vulnerability note reports:

An Integer overflow in the LibTIFF library may allow a remote attacker to cause a divide-by-zero error that results in a denial-of-service condition.

infamous41md reports:

zgv uses malloc() frequently to allocate memory for storing image data. When calculating how much to allocate, user supplied data from image headers is multiplied and/or added without any checks for arithmetic overflows. We can overflow numerous calculations, and cause small buffers to be allocated. Then we can overflow the buffer, and eventually execute code. There are a total of 11 overflows that are exploitable to execute arbitrary code.

These bugs exist in both zgv and xzgv.

In a Mozilla bug report, Daniel Kleinsinger writes:

I was comparing treatment of attachments opened directly from emails on different platforms. I discovered that Linux builds save attachments in /tmp with world readable rights. This doesn't seem like a good thing. Couldn't someone else logged onto the same machine read your attachments?

This could expose the contents of downloaded files or email attachments to other users on a multi-user system.

An iDEFENSE Security Advisory reports:

Remote exploitation of an input validation vulnerability in AWStats allows attackers to execute arbitrary commands under the privileges of the web server.

The problem specifically exists when the application is running as a CGI script on a web server. The "configdir" parameter contains unfiltered user-supplied data that is utilized in a call to the Perl routine open()...

Successful exploitation allows remote attackers to execute arbitrary commands under the privileges of the web server. This can lead to further compromise as it provides remote attackers with local access.

An iDEFENSE Security Advisory reports:

Remote exploitation of a buffer overflow vulnerability in The ImageMagick's Project's ImageMagick PSD image-decoding module could allow an attacker to execute arbitrary code.

Exploitation may allow attackers to run arbitrary code on a victim's computer if the victim opens a specially formatted image. Such images could be delivered by e-mail or HTML, in some cases, and would likely not raise suspicion on the victim's part. Exploitation is also possible when a web-based application uses ImageMagick to process user-uploaded image files.

D. J. Bernstein reports that Bartlomiej Sieka has discovered several security vulnerabilities in lppasswd, which is part of CUPS. In the following excerpt from Bernstein's email, CVE names have been added for each issue:

First, lppasswd blithely ignores write errors in fputs(line,outfile) at lines 311 and 315 of lppasswd.c, and in fprintf(...) at line 346. An attacker who fills up the disk at the right moment can arrange for /usr/local/etc/cups/passwd to be truncated. (CAN-2004-1268)

Second, if lppasswd bumps into a file-size resource limit while writing passwd.new, it leaves passwd.new in place, disabling all subsequent invocations of lppasswd. Any local user can thus disable lppasswd... (CAN-2004-1269)

Third, line 306 of lppasswd.c prints an error message to stderr but does not exit. This is not a problem on systems that ensure that file descriptors 0, 1, and 2 are open for setuid programs, but it is a problem on other systems; lppasswd does not check that passwd.new is different from stderr, so it ends up writing a user-controlled error message to passwd if the user closes file descriptor 2. (CAN-2004-1270)

Note: The third issue, CAN-2004-1270, does not affect FreeBSD 4.6-RELEASE or later systems, as these systems ensure that the file descriptors 0, 1, and 2 are always open for set-user-ID and set-group-ID programs.

Ariel Berkman has discovered a buffer overflow vulnerability in CUPS's HPGL input driver. This vulnerability could be exploited to execute arbitrary code with the permission of the CUPS server by printing a specially crated HPGL file.

The Debian Security Team reports:

Javier Fernández-Sanguino Peña from the Debian Security Audit Project discovered a temporary file vulnerability in the mysqlaccess script of MySQL that could allow an unprivileged user to let root overwrite arbitrary files via a symlink attack and could also could unveil the contents of a temporary file which might contain sensitive information.

Yosef Klein and Limin Wang have found a buffer overflow vulnerability in unrtf that can allow an attacker to execute arbitrary code with the permissions of the user running unrtf, by running unrtf on a specially crafted rtf document.

Maurycy Prodeus reports a critical vulnerability in Mozilla-based browsers:

Mozilla browser supports NNTP urls. Remote side is able to trigger news:// connection to any server. I found a flaw in NNTP handling code which may cause heap overflow and allow remote attacker to execute arbitrary code on client machine.

Yuri D'Elia has found a buffer overflow vulnerability in mpg123's parsing of frame headers in input streams. This vulnerability can potentially lead to execution of arbitrary code with the permissions of the user running mpg123, if the user runs mpg123 on a specially crafted MP2 or MP3 file.

The squid patches page notes:

WCCP_I_SEE_YOU messages contain a 'number of caches' field which should be between 1 and 32. Values outside that range may crash Squid if WCCP is enabled, and if an attacker can spoof UDP packets with the WCCP router's IP address.

The squid patches page notes:

A malicious gopher server may return a response with very long lines that cause a buffer overflow in Squid.

Workaround: Since gopher is very obscure these days, do not allow Squid to any gopher servers. Use an ACL rule like:

acl Gopher proto gopher
http_access deny Gopher

A xine security announcement states:

A heap overflow has been found in the DVD subpicture decoder of xine-lib. This can be used for a remote heap overflow exploit, which can, on some systems, lead to or help in executing malicious code with the permissions of the user running a xine-lib based media application.

A xine security announcement states:

Several string overflows on the stack have been fixed in xine-lib, some of them can be used for remote buffer overflow exploits leading to the execution of arbitrary code with the permissions of the user running a xine-lib based media application.

Stack-based string overflows have been found:

1. in the code which handles VideoCD MRLs
2. in VideoCD code reading the disc label
3. in the code which parses text subtitles and prepares them for display

A xine security announcement states:

Multiple vulnerabilities have been found and fixed in the Real-Time Streaming Protocol (RTSP) client for RealNetworks servers, including a series of potentially remotely exploitable buffer overflows. This is a joint advisory by the MPlayer and xine teams as the code in question is common to these projects.

Severity: High (arbitrary remote code execution under the user ID running the player) when playing Real RTSP streams. At this time, there is no known exploit for these vulnerabilities.

A flaw in HylaFAX may allow an attacker to bypass normal authentication by spoofing their DNS PTR records.

Steve Kemp has found buffer overflows in the handling of the command line flag -KCONV and the XSHISENLIB environment variable. Ulf Härnhammer has detected an unbounded copy from the GECOS field to a char array. All overflows can be exploited to gain group games privileges.

The setuid root elvprsv utility, used to preserve recovery helvis files, can be abused by local users to delete with root privileges.

The problem is that elvprsv deletes files when it thinks they have become corrupt. When elvprsv is pointed to a normal file then it will almost always think the file is corrupt and deletes it. This behavior may be exploited by local attackers to delete critical files.

Once a recovery file has been preserved by the setuid root elvprsv utility it is placed in a worldreadable directory with worldreadable permissions. This possibly allows sensitive information to leak.

In addition to this information leak, it is possible for users to recover files that belong to other users by using elvrec, another setuid root binary.

dillo contains a format string vulnerability which could lead to execution of arbitrary code simply by viewing a web page or opening a HTML file.

When downloading a batch of files from an FTP server the mget command does not check for directory escapes. A specially crafted file on the FTP server could then potentially overwrite an existing file of the user.

Dmitry V. Levin found a potential integer overflow in the tiffdump utility which could lead to execution of arbitrary code. This could be exploited by tricking an user into executing tiffdump on a specially crafted tiff image.

In an iDEFENSE Security Advisory infamous41md reports:

Remote exploitation of a heap-based buffer overflow vulnerability within the LibTIFF package could allow attackers to execute arbitrary code.

The vulnerability specifically exists due to insufficient validation of user-supplied data when calculating the size of a directory entry. A TIFF file includes a number of directory entry header fields that describe the data in the file. Included in these entries is an entry count and offset value that are calculated to determine the size and location of the data for that entry.

Ciaran McCreesh discovered news ways in which a VIM modeline can be used to trojan a text file. The patch by Bram Moolenaar reads:

Problem: Unusual characters in an option value may cause unexpected behavior, especially for a modeline. (Ciaran McCreesh)

Solution: Don't allow setting termcap options or 'printdevice' or 'titleold' in a modeline. Don't list options for "termcap" and "all" in a modeline. Don't allow unusual characters in 'filetype', 'syntax', 'backupext', 'keymap', 'patchmode' and 'langmenu'.

Note: It is generally recommended that VIM users use set nomodeline in ~/.vimrc to avoid the possibility of trojaned text files.

Danny Lungstrom has found two buffer overflow vulnerabilities in pcal which can lead to execution of arbitrary code by making a user run pcal on a specially crafted calendar file.

1. The function host_aton() can overflow a buffer if it is presented with an illegal IPv6 address that has more than 8 components.

2. The second report described a buffer overflow in the function spa_base64_to_bits(), which is part of the code for SPA authentication.

A buffer overflow vulnerability exists in the playlist processing of mpg123. A specially crafted playlist entry can cause a stack overflow that can be used to inject arbitrary code into the mpg123 process

Note that a malicious playlist, demonstrating this vulnerability, was released by the bug finder and may be used as a template by attackers.

A buffer overflow vulnerability has been detected in the greed URL handling code. This bug can especially be a problem when greed is used to process GRX (GetRight) files that originate from untrusted sources.

The bug finder, Manigandan Radhakrishnan, gave the following description:

Here are the bugs. First, in main.c, DownloadLoop() uses strcat() to copy an input filename to the end of a 128-byte COMMAND array. Second, DownloadLoop() passes the input filename to system() without checking for special characters such as semicolons.

Two buffer overflow vulnerabilities where detected. Both issues can be used by local users to gain group games privileges on affected systems.

The first overflow exists in the map name handling and can be triggered when a very long name is given to the program during command-line execution

The second overflow exists in the username processing while writing the players score to disk. Excessivly long usernames, set via the USER environment variable, are stored without any length checks in a memory buffer.

Timo Sirainen reports:

There are various bugs in up-imapproxy which can crash it. Since up-imapproxy runs in a single process with each connection handled in a separate thread, any crash kills all the connections and stops listening for new ones.

In 64bit systems it might be possible to make it leak data (mails, passwords, ..) from other connections to attacker's connection. However I don't think up-imapproxy actually works in any 64bit system so this is just a theoretical problem.

Albert Puigsech Galicia reports that Konqueror (more specifically kio_ftp) and Microsoft Internet Explorer are vulnerable to a FTP command injection vulnerability which can be exploited by tricking an user into clicking a specially crafted FTP URI.

It is also reported by Ian Gulliver and Emanuele Balla that this vulnerability can be used to tricking a client into sending out emails without user interaction.

A Secunia Security Advisory reports that Javier Fernández-Sanguino Peña has found temporary file creation vulnerabilities in the fixps and psmandup scripts which are part of a2ps. These vulnerabilities could lead to an attacker overwriting arbitrary files with the credentials of the user running the vulnerable scripts.

Due to a buffer overflow in the open_aiff_file function in demux_aiff.c, a remote attacker is able to execute arbitrary code via a modified AIFF file.

José Antonio Calvo discovered a bug in the Jabber 1.x server. According to Matthias Wimmer:

Without this patch, it is possible to remotly crash jabberd14, if there is access to one of the following types of network sockets:

• Socket accepting client connections
• Socket accepting connections from other servers
• Socket connecting to an other Jabber server
• Socket accepting connections from server components
• Socket connecting to server components

This is any socket on which the jabberd server parses XML!

The problem existed in the included expat XML parser code. This patch removes the included expat code from jabberd14 and links jabberd against an installed version of expat.

Applying an empty ACL list results in unexpected behavior: anything will match an empty ACL list. For example,

The meaning of the configuration gets very confusing when we encounter empty ACLs such as

acl something src "/path/to/empty_file.txt"
http_access allow something somewhere

gets parsed (with warnings) as

http_access allow somwhere

And similarily if you are using proxy_auth acls without having any auth schemes defined.

An Ethreal Security Advisories reports:

Issues have been discovered in the following protocol dissectors:

• Matthew Bing discovered a bug in DICOM dissection that could make Ethereal crash.
• An invalid RTP timestamp could make Ethereal hang and create a large temporary file, possibly filling available disk space.
• The HTTP dissector could access previously-freed memory, causing a crash.
• Brian Caswell discovered that an improperly formatted SMB packet could make Ethereal hang, maximizing CPU utilization.

Impact: It may be possible to make Ethereal crash or run arbitrary code by injecting a purposefully malformed packet onto the wire or by convincing someone to read a malformed packet trace file.

An iDEFENSE Security Advisory reports:

Remote exploitation of a buffer overflow vulnerability in the xpdf PDF viewer, as included in multiple Linux distributions, could allow attackers to execute arbitrary code as the user viewing a PDF file. The offending code can be found in the Gfx::doImage() function in the source file xpdf/Gfx.cc.

An iDEFENSE Security Advisory reports:

Remote exploitation of a buffer overflow in version 5.09 of Adobe Acrobat Reader for Unix could allow for execution of arbitrary code.

The vulnerability specifically exists in a the function mailListIsPdf(). This function checks if the input file is an email message containing a PDF. It unsafely copies user supplied data using strcat into a fixed sized buffer.

A Debian security advisory reports:

A problem has been discovered in ecartis, a mailing-list manager, which allows an attacker in the same domain as the list admin to gain administrator privileges and alter list settings.

iDEFENSE and the MPlayer Team have found multiple vulnerabilities in MPlayer:

• Potential heap overflow in Real RTSP streaming code
• Potential stack overflow in MMST streaming code
• Multiple buffer overflows in BMP demuxer
• Potential heap overflow in pnm streaming code
• Potential buffer overflow in mp3lib

These vulnerabilities could allow a remote attacker to execute arbitrary code as the user running MPlayer. The problem in the pnm streaming code also affects xine.

A MIT krb5 Security Advisory reports:

The MIT Kerberos 5 administration library (libkadm5srv) contains a heap buffer overflow in password history handling code which could be exploited to execute arbitrary code on a Key Distribution Center (KDC) host. The overflow occurs during a password change of a principal with a certain password history state. An administrator must have performed a certain password policy change in order to create the vulnerable state.

An authenticated user, not necessarily one with administrative privileges, could execute arbitrary code on the KDC host, compromising an entire Kerberos realm.

Greg MacManus, iDEFENSE Labs reports:

Remote exploitation of an integer overflow vulnerability in the smbd daemon included in Samba 2.0.x, Samba 2.2.x, and Samba 3.0.x prior to and including 3.0.9 could allow an attacker to cause controllable heap corruption, leading to execution of arbitrary commands with root privileges.

Successful remote exploitation allows an attacker to gain root privileges on a vulnerable system. In order to exploit this vulnerability an attacker must possess credentials that allow access to a share on the Samba server. Unsuccessful exploitation attempts will cause the process serving the request to crash with signal 11, and may leave evidence of an attack in logs.

Secunia reports:

Multiple vulnerabilities have been reported in PHP, which can be exploited to gain escalated privileges, bypass certain security restrictions, gain knowledge of sensitive information, or compromise a vulnerable system.

When a user is granted access to a database with a name containing an underscore and the underscore is not escaped then that user might also be able to access other, similarly named, databases on the affected system.

The problem is that the underscore is seen as a wildcard by MySQL and therefore it is possible that an admin might accidently GRANT a user access to multiple databases.

Dean Ellis reported a denial of service vulnerability in the MySQL server:

Multiple threads ALTERing the same (or different) MERGE tables to change the UNION eventually crash the server or hang the individual threads.

Note that a script demonstrating the problem is included in the MySQL bug report. Attackers that have control of a MySQL account can easily use a modified version of that script during an attack.

A special crafted MySQL FTS request can cause the server to crash. Malicious MySQL users can abuse this bug in a denial of service attack against systems running an affected MySQL daemon.

Note that because this bug is related to the parsing of requests, it may happen that this bug is triggered accidently by a user when he or she makes a typo.

The mysql_real_connect function doesn't properly handle DNS replies by copying the IP address into a buffer without any length checking. A specially crafted DNS reply may therefore be used to cause a buffer overflow on affected systems.

Note that whether this issue can be exploitable depends on the system library responsible for the gethostbyname function. The bug finder, Lukasz Wojtow, explaines this with the following words:

In glibc there is a limitation for an IP address to have only 4 bytes (obviously), but generally speaking the length of the address comes with a response for dns query (i know it sounds funny but read rfc1035 if you don't believe). This bug can occur on libraries where gethostbyname function takes length from dns's response

A Red Hat advisory reports:

Oleksandr Byelkin discovered that "ALTER TABLE ... RENAME" checked the CREATE/INSERT rights of the old table instead of the new one.

Table access restrictions, on the affected MySQL servers, may accidently or intentially be bypassed due to this bug.

A phpMyAdmin security announcement reports:

Command execution: since phpMyAdmin 2.6.0-pl2, on a system where external MIME-based transformations are activated, an attacker can put into MySQL data an offensive value that starts a shell command when browsed.

Enabling PHP safe mode on the server can be used as a workaround for this vulnerability.

A phpMyAdmin security announcement reports:

File disclosure: on systems where the UploadDir mecanism is active, read_dump.php can be called with a crafted form; using the fact that the sql_localfile variable is not sanitized can lead to a file disclosure.

Enabling PHP safe mode on the server can be used as a workaround for this vulnerability.

Jan Minar reports that there exists multiple vulnerabilities in wget:

Wget erroneously thinks that the current directory is a fair game, and will happily write in any file in and below it. Malicious HTTP response or malicious HTML file can redirect wget to a file that is vital to the system, and wget will create/append/overwrite it.

Wget apparently has at least two methods of ``sanitizing'' the potentially malicious data it receives from the HTTP stream, therefore a malicious redirects can pass the check. We haven't find a way to trick wget into writing above the parent directory, which doesn't mean it's not possible.

Malicious HTTP response can overwrite parts of the terminal so that the user will not notice anything wrong, or will believe the error was not fatal.

When browsing SMB shares with Konqueror, shares with authentication show up with hidden password in the browser bar. It is possible to store the URL as a shortcut on the desktop where the password is then available in plain text.

A malformed Referer header field causes the Apache ap_parse_uri_components function to discard it with the result that a pointer is not initialized. The mod_access_referer module does not take this into account with the result that it may use such a pointer.

The null pointer vulnerability may possibly be used in a remote denial of service attack against affected Apache servers.

The squid-2.5 patches pages notes:

In certain conditions Squid returns random data as error messages in response to malformed host name, possibly leaking random internal information which may come from other requests.

The hide_cvsroot and forbidden configuration options are not properly honored by viewcvs when exporting to a tar file which can lead to information leakage.

cscope is vulnerable to a symlink attack which could lead to an attacker overwriting arbitrary files with the permissions of the user running cscope.

A LSS Security Advisory reports:

There is a buffer overflow vulnerability in getnickuserhost() function that is called when BNC is processing response from IRC server.

Vulnerability can be exploited if attacker tricks user to connect to his fake IRC server that will exploit this vulnerability. If the attacker has access to BNC proxy server, this vulnerability can be used to gain shell access on machine where BNC proxy server is set.

Jason Wies identified both rssh & scponly have a vulnerability that allows arbitrary command execution. He reports:

The problem is compounded when you recognize that the main use of rssh and scponly is to allow file transfers, which in turn allows a malicious user to transfer and execute entire custom scripts on the remote machine.

The environment variable HOME is copied without regard to buffer size, which can be used to gain elevated privilege if the binary is installed setgid games, and a string is read from the high score file without bounds check.

The port installs the binary without setgid, but with a world-writable high score file.

A HexView security advisory reports:

When zip performs recursive folder compression, it does not check for the length of resulting path. If the path is too long, a buffer overflow occurs leading to stack corruption and segmentation fault. It is possible to exploit this vulnerability by embedding a shellcode in directory or file name. While the issue is not of primary concern for regular users, it can be critical for environments where zip archives are re-compressed automatically using Info-Zip application.

If non-root access is enabled in sudoscript, any member of the ssers group can send a SIGHUP signal to any process.

Caused by improper bounds-checking of username and password in the C2S module, it is possible for an attacker to cause a remote buffer overflow. The server directly handles the userinput with SQL backend functions - malicious input may lead to buffer overflow.

Donato Ferrante reported an exploitable buffer overflow in this software package. Any user that can login with 'admin' privileges can abuse it, trough the $RedirectAll command, to execute arbitrary code.

Ludwig Nussel has discovered a buffer overflow vulnerability in unarj's handling of long filenames which could potentially lead to execution of arbitrary code with the permissions of the user running unarj.

unarj has insufficient checks for filenames that contain ... This can allow an attacker to overwrite arbitrary files with the permissions of the user running unarj.

The Sun Java Plugin capability in Java 2 Runtime Environment (JRE) 1.4.2_01, 1.4.2_04, and possibly earlier versions, does not properly restrict access between Javascript and Java applets during data transfer, which allows remote attackers to load unsafe classes and execute arbitrary code.

Buffer overflow vulnerabilities have been reported to exist in this software package. The vulnerabilities can be triggered by a remote server and can be used to inject malicious code in the ProZilla process.

To support MULTIAPPENDS the cmd_append handler uses the global stage array. This array is one of the things that gets destructed when the fatal() function is triggered. When the Cyrus IMAP code adds new entries to this array this is done with the help of the postfix increment operator in combination with memory allocation functions. The increment is performed on a global variable counting the number of allocated stages. Because the memory allocation function can fail and therefore internally call fatal() this construct is undefined arcording to ANSI C. This means that it is not clearly defined if the numstage counter is already increased when fatal() is called or not. While older gcc versions increase the counter after the memory allocation function has returned, on newer gcc versions (3.x) the counter gets actually increased before. In such a case the stage destructing process will try to free an uninitialised and maybe attacker supplied pointer. Which again could lead to remote code execution. (Because it is hard for an attacker to let the memory allocation functions fail in the right moment no PoC code for this problem was designed)

The argument parser of the fetch command suffers a bug very similiar to the partial command problem. Arguments like "body[p", "binary[p" or "binary[p" will be wrongly detected and the bufferposition can point outside of the allocated buffer for the rest of the parsing process. When the parser triggers the PARSE_PARTIAL macro after such a malformed argument was received this can lead to a similiar one byte memory corruption and allows remote code execution, when the heap layout was successfully controlled by the attacker.

Due to a bug within the argument parser of the partial command an argument like "body[p" will be wrongly detected as "body.peek". Because of this the bufferposition gets increased by 10 instead of 5 and could therefore point outside the allocated memory buffer for the rest of the parsing process. In imapd versions prior to 2.2.7 the handling of "body" or "bodypeek" arguments was broken so that the terminating ']' got overwritten by a '\0'. Combined the two problems allow a potential attacker to overwrite a single byte of malloc() control structures, which leads to remote code execution if the attacker successfully controls the heap layout.

When the option imapmagicplus is activated on a server the PROXY and LOGIN commands suffer a standard stack overflow, because the username is not checked against a maximum length when it is copied into a temporary stack buffer. This bug is especially dangerous because it can be triggered before any kind of authentification took place.

Multiple cross-site scripting vulnerabilities, caused by improper input parameter sanitizing, were detected in phpMyAdmin, which may enable an attacker to do cross-site scripting attacks.

An integer overflow condition in fetch(1) in the processing of HTTP headers can result in a buffer overflow.

A malicious server or CGI script can respond to an HTTP or HTTPS request in such a manner as to cause arbitrary portions of the client's memory to be overwritten, allowing for arbitrary code execution.

Caused by improper bounds checking of certain trans2 requests, there is a possible buffer overrun in smbd. The attacker needs to be able to create files with very specific Unicode filenames on the share to take advantage of this issue.

Hans Ulrich Niedermann reports:

The TWiki search function uses a user supplied search string to compose a command line executed by the Perl backtick (``) operator.

The search string is not checked properly for shell metacharacters and is thus vulnerable to search string containing quotes and shell commands.

IMPACT: An attacker is able to execute arbitrary shell commands with the privileges of the TWiki process.

A Gentoo Linux Security Advisory reports:

Florian Schilhabel of the Gentoo Linux Security Audit project found a format string vulnerability in Proxytunnel. When the program is started in daemon mode (-a [port]), it improperly logs invalid proxy answers to syslog.

A malicious remote server could send specially-crafted invalid answers to exploit the format string vulnerability, potentially allowing the execution of arbitrary code on the tunnelling host with the rights of the Proxytunnel process.

A Sudo Security Alerts reports:

A flaw in exists in sudo's environment sanitizing prior to sudo version 1.6.8p2 that could allow a malicious user with permission to run a shell script that utilized the bash shell to run arbitrary commands.

The Ruby CGI.rb module contains a bug which can cause the CGI module to go into an infinite loop, thereby causing a denial-of-service situation on the web server by using all available CPU time.

Karol Wiesek at iDEFENSE reports:

A remote attacker could cause an smbd process to consume abnormal amounts of system resources due to an input validation error when matching filenames containing wildcard characters.

Although samba.org classifies this as a DoS vulnerability, several members of the security community believe it may be exploitable for arbitrary code execution.

Gnats suffers from a format string bug, which may enable an attacker to execute arbitary code.

A SquirrelMail Security Notice reports:

There is a cross site scripting issue in the decoding of encoded text in certain headers. SquirrelMail correctly decodes the specially crafted header, but doesn't sanitize the decoded strings.

A siyahsapka.org advisory reads:

Hafiye-1.0 doesnt filter the payload when printing it to the terminal. A malicious attacker can send packets with escape sequence payloads to exploit this vulnerability.

If Hafiye has been started with -n packet count option , the vulnerability could allow remote code execution. For remote code execution the victim must press Enter after program exit.

Note that it appears that this bug can only be exploited in conjunction with a terminal emulator that honors the appropriate escape sequences.

Data supplied by a remote server is used as the format string instead of as parameters in a syslog() call. This may lead to crashes or potential running of arbitrary code. It is only a problem when running in daemon mode (very common) and when using some service types.

There exists a buffer overflow vulnerability in ImageMagick's EXIF parsing code which may lead to execution of arbitrary code.

It is possible for remote attackers to cause a denial-of-service scenario on Apache 2.0.52 and earlier by sending an HTTP GET request with a MIME header containing multiple lines full of whitespaces.

Socat Security Advisory 1 states:

socat up to version 1.4.0.2 contains a syslog() based format string vulnerability. This issue was originally reported by CoKi on 19 Oct.2004 http://www.nosystem.com.ar/advisories/advisory-07.txt. Further investigation showed that this vulnerability could under some circumstances lead to local or remote execution of arbitrary code with the privileges of the socat process.

infamous41md reports that libxml contains multiple buffer overflows in the URL parsing and DNS name resolving functions. These vulnerabilities could lead to execution of arbitrary code.

An AMaViS Security Announcement reports that a vulnerability exist in the Archive::Zip Perl module which may lead to bypass of malicious code in anti-virus programs by creating specially crafted ZIP files.

There is a buffer overflow in a function used by mod_include that may enable a local user to gain privileges of a httpd child. Only users that are able to create SSI documents can take advantage of that vulnerability.

The make_oidjoins_check script in the PostgreSQL RDBMS has insecure handling of temporary files, which could lead to an attacker overwriting arbitrary files with the credentials of the user running the make_oidjoins_check script.

infamous41md reports about the GD Graphics Library:

There is an integer overflow when allocating memory in the routine that handles loading PNG image files. This later leads to heap data structures being overwritten. If an attacker tricked a user into loading a malicious PNG image, they could leverage this into executing arbitrary code in the context of the user opening image.

There is a bug in SSH2 support that allows a server to execute malicious code on a connecting PuTTY client. This attack can be performed before host key verification happens, so a different machine -- man in the middle attack -- could fake the machine you are connecting to.

wzdftpd contains a potential remote Denial-of-Service.

There is a format string bug in rssh that enables an attacker to execute arbitrary code from an account configured to use rssh. On FreeBSD it is only possible to compromise the rssh running account, not root.

A Horde Team announcement states that a potential cross-site scripting vulnerability in the help window has been corrected. The vulnerability appears to involve the handling of the topic and module parameters of the help window template.

The bogofilter team has been provided with a test case of a malformatted (non-conformant) RFC-2047 encoded word that can cause bogofilter versions 0.92.7 and prior to try to write a NUL byte into a memory location that is either one byte past the end of a flex buffer or to a location that is the negative of the encoded word's start of payload data, causing a segmentation fault.

Chris Evans discovered several integer arithmetic overflows in the xpdf 2 and xpdf 3 code bases. The flaws have impacts ranging from denial-of-service to arbitrary code execution.

The Gaim team discovered denial-of-service vulnerabilities in the MSN protocol handler:

After accepting a file transfer request, Gaim will attempt to allocate a buffer of a size equal to the entire filesize, this allocation attempt will cause Gaim to crash if the size exceeds the amount of available memory.

Gaim allocates a buffer for the payload of each message received based on the size field in the header of the message. A malicious peer could specify an invalid size that exceeds the amount of available memory.

Sean infamous42md reports:

When a remote server provides a large "content-length" header value, Gaim will attempt to allocate a buffer to store the content, however this allocation attempt will cause Gaim to crash if the length exceeds the amount of possible memory. This happens when reading profile information on some protocols. It also happens when smiley themes are installed via drag and drop.

Sean infamous42md reports several situations in gaim that may result in exploitable buffer overflows:

• Rich Text Format (RTF) messages in Novell GroupWise protocol
• Unsafe use of gethostbyname in zephyr protocol
• URLs which are over 2048 bytes long once decoded

Sean infamous42md reports that a malicous GroupWise messaging server may be able to exploit a heap buffer overflow in gaim, leading to arbitrary code execution.

The Gaim Security Issues page documents a problem with installing smiley themes from an untrusted source:

To install a new smiley theme, a user can drag a tarball from a graphical file manager, or a hypertext link to one from a web browser. When a tarball is dragged, Gaim executes a shell command to untar it. However, it does not escape the filename before sending it to the shell. Thus, a specially crafted filename could execute arbitrary commands if the user could be convinced to drag a file into the smiley theme selector.

Due to a buffer overflow in the MSN protocol support for gaim 0.79 to 1.0.1, it is possible for remote clients to do a denial-of-service attack on the application. This is caused by an unbounded copy operation, which writes to the wrong buffer.

It is possible for clients to use any cipher suite configured by the virtual host, whether or not a certain cipher suite is selected for a specific directory. This might result in clients using a weaker encryption than originally configured.

Carlos Barros reports that mpg123 contains two buffer overflows. These vulnerabilities can potentially lead to execution of arbitrary code.

The first buffer overflow can occur when mpg123 parses a URL with a user-name/password field that is more than 256 characters long. This problem can be triggered either locally or remotely via a specially crafted play list. The second potential buffer overflow may be triggered locally by a specially crafted symlink to the mpg123 binary. This problem is not as serious, since mpg123 is not installed setuid by default.

The Apache HTTP Server 2.0.51 release notes report that the following issues have been fixed:

A segfault in mod_ssl which can be triggered by a malicious remote server, if proxying to SSL servers has been configured. [CAN-2004-0751]

A potential infinite loop in mod_ssl which could be triggered given particular timing of a connection abort. [CAN-2004-0748]

From the phpMyAdmin 2.6.0p2 release notes:

If PHP is not running in safe mode, a problem in the MIME-based transformation system (with an "external" transformation) allows to execute any command with the privileges of the web server's user.

cabextract has insufficient checks for file names that contain ../. This can cause files to be extracted to the parent directory.

Rudolf Polzer reports:

a2ps builds a command line for file() containing an unescaped version of the file name, thus might call external programs described by the file name. Running a cronjob over a public writable directory a2ps-ing all files in it - or simply typing "a2ps *.txt" in /tmp - is therefore dangerous.

Niels Heinen reports that ifmail allows one to specify a configuration file. Since ifmail runs set-user-ID `news', this may allow a local attacker to write to arbitrary files or execute arbitrary commands as the `news' user.

A Computer Academic Underground advisory describes the consequences of imwheel's handling of the process ID file (PID file):

imwheel exclusively uses a predictably named PID file for management of multiple imwheel processes. A race condition exists when the -k command-line option is used to kill existing imwheel processes. This race condition may be used by a local user to Denial of Service another user using imwheel, lead to resource exhaustion of the host system, or append data to arbitrary files.

A remote attacker is able to cause a denial-of-service situation, when NTLM authentication is enabled in squid. NTLM authentication uses two functions which lack correct offset checking.

Fernando Quintero reports that Cacti 0.8.5a suffers from a SQL injection attack where an attacker can change the password for any Cacti user. This attack is not possible if the PHP option magic_quotes_gpc is set to On, which is the default for PHP in FreeBSD.

A OpenPKG Security Advisory reports:

Triggered by a report to Packet Storm from Virulent, a format string vulnerability was found in mod_ssl, the Apache SSL/TLS interface to OpenSSL, version (up to and including) 2.8.18 for Apache 1.3. The mod_ssl in Apache 2.x is not affected. The vulnerability could be exploitable if Apache is used as a proxy for HTTPS URLs and the attacker established a own specially prepared DNS and origin server environment.

Tor has various remote crashes which could lead to a remote denial-of-service and be used to defeat clients anonymity. It is not expected that these vulnerabilities are exploitable for arbitrary code execution.

Caused by improper filtering of HTML code in the status display, it is possible for a remote user to execute scripting code in the target user's browser.

It is possible to execute remote code simply using HTTP request plus 31 headers followed by a shellcode that will be executed directly.

A remote attacker may be able to crash the freeRADIUS Server due to three independant bugs in the function which does improper checking values while processing RADIUS attributes.

Amit Klein reports about Xerces-C++:

An attacker can craft a malicious XML document, which uses XML attributes in a way that inflicts a denial of service condition on the target machine (XML parser). The result of this attack is that the XML parser consumes all the CPU.

Pages in the administration panel of Wordpress are vulnerable for XSS attacks.

Dmitry V. Levin discovered numerous integer overflow bugs in libtiff. Most of these bugs are related to memory management, and are believed to be exploitable for arbitrary code execution.

Certain methods of authenticated remote printing in CUPS can disclose user names and passwords in the log files.

A workaround for this problem is to set more strict access permissions on the CUPS logfiles.

The audio player Zinf is vulnerable to a buffer-overflow bug in the management of the playlist files.

Chris Evans discovered several heap buffer overflows in libtiff's RLE decoder. These overflows could be triggered by a specially-crafted TIFF image file, resulting in an application crash and possibly arbitrary code execution.

From Gentoo advisory GLSA 200410-01:

sharutils contains two buffer overflows. Ulf Harnhammar discovered a buffer overflow in shar.c, where the length of data returned by the wc command is not checked. Florian Schilhabel discovered another buffer overflow in unshar.c.

An attacker could exploit these vulnerabilities to execute arbitrary code as the user running one of the sharutils programs.

Caused by an untested return value, and a resulting null pointer dereference, it is possible for an attacker to crash the application. However, the attacker must first hijack the connection between Mail Notification and the Gmail or IMAP server.

The Squid-2.5 patches page notes:

If a certain malformed SNMP request is received squid restarts with a Segmentation Fault error.

This only affects squid installations where SNMP is explicitly enabled via "make config". As a workaround, SNMP can be disabled by defining "snmp_port 0" in squid.conf.

The Cyrus SASL DIGEST-MD5 plugin contains a potential buffer overflow when quoting is required in the output.

The Cyrus SASL library, libsasl, contains functions which may load dynamic libraries. These libraries may be loaded from the path specified by the environmental variable SASL_PATH, which in some situations may be fully controlled by a local attacker. Thus, if a set-user-ID application (such as chsh) utilizes libsasl, it may be possible for a local attacker to gain superuser privileges.

The script vulnerabilities can only be exposed with certain browsers and allow XSS attacks when viewing HTML messages with the HTML MIME viewer

Jon Nistor reported that the FreeBSD port of bmon was installed set-user-ID root, and executes commands using relative paths. This could allow local user to easily obtain root privileges.

Patric Hornik reports on a problem in the certificate chain verification procedures of GnuTLS that may result in a denial-of-service vulnerability:

The certificate chain should be verified from last root certificate to the first certificate. Otherwise a lot of unauthorized CPU processing can be forced to check certificate signatures signed with arbitrary RSA/DSA keys chosen by attacker.

In GnuTLS the signatures are checked from first to last certificate, there is no limit on size of keys and no limit on length of certificate chain.

Stefano Di Paola discovered an issue with PHP that could allow someone to upload a file to any directory writeable by the httpd process. Any sanitizing performed on the prepended directory path is ignored. This bug can only be triggered if the $_FILES element name contains an underscore.

Stefano Di Paola reports:

Bad array parsing in php_variables.c could lead to show arbitrary memory content such as pieces of php code and other data. This affects all GET, POST or COOKIES variables.

In a Bugtraq posting, infamous41md(at)hotpop.com reported:

there are at least 5 exploitable buffer and heap overflows in the image handling code. this allows someone to craft a malicious image, trick a user into viewing the file in xv, and upon viewing that image execute arbitrary code under privileges of the user viewing image. note the AT LEAST part of the above sentence. there is such a plethora of bad code that I just stopped reading after a while. there are at least 100 calls to sprintf() and strcpy() with no regards for bounds of buffers. 95% of these deal with program arguments or filenames, so they are of no interest to exploit. however I just got sick of reading this code after not too long. so im sure there are still other overflows in the image handling code for other image types.

The posting also included an exploit.

David Watson reports a symlink vulnerability in getmail. If run as root (not the recommended mode of operation), a local user may be able to cause getmail to write files in arbitrary directories via a symlink attack on subdirectories of the maildir.

The syscons CONS_SCRSHOT ioctl(2) does insufficient validation of its input arguments. In particular, negative coordinates or large coordinates may cause unexpected behavior.

It may be possible to cause the CONS_SCRSHOT ioctl to return portions of kernel memory. Such memory might contain sensitive information, such as portions of the file cache or terminal buffers. This information might be directly useful, or it might be leveraged to obtain elevated privileges in some way. For example, a terminal buffer might include a user-entered password.

This bug may be exploitable by users who have access to the physical console or can otherwise open a /dev/ttyv* device node.

Thomas Walpuski noted when OpenSSL would detect an error condition for a peer certificate, racoon mistakenly ignored the error. This could allow five invalid certificate states to properly be used for authentication.

Fix bug that might cause IP-based access control rules not to be interpreted correctly on 64-bit platforms.

Several scripting vulnerabilities were discovered and corrected in Mozilla:

CAN-2004-0905

javascript; links dragged onto another frame or page allows an attacker to steal or modify sensitive information from other sites. The user could be convinced to drag obscurred links in the context of a game or even a fake scrollbar. If the user could be convinced to drag two links in sequence into a separate window (not frame) the attacker would be able to run arbitrary programs.

CAN-2004-0908

Untrusted javascript code can read and write to the clipboard, stealing any sensitive data the user might have copied. Workaround: disable javascript

CAN-2004-0909

Signed scripts requesting enhanced abilities could construct the request in a way that led to a confusing grant dialog, possibly fooling the user into thinking the privilege requested was inconsequential while actually obtaining explicit permission to run and install software. Workaround: Never grant enhanced abilities of any kind to untrusted web pages.

According to the Mozilla project:

An attacker who could lure users into clicking in particular places, or typing specific text, could cause a security permission or software installation dialog to pop up under the user's mouse click, clicking on the grant (or install) button.

When processing URIs that contain an unqualified host name-- specifically, a domain name of only one component-- Mozilla will perform matching against the first component of the domain name in SSL certificates. In other words, in some situations, a certificate issued to "www.example.com" will be accepted as matching "www".

According to a Samba Team security notice:

A security vulnerability has been located in Samba 2.2.x <= 2.2.11 and Samba 3.0.x <= 3.0.5. A remote attacker may be able to gain access to files which exist outside of the share's defined path. Such files must still be readable by the account used for the connection.

The original notice for CAN-2004-0815 indicated that Samba 3.0.x <= 3.0.5 was vulnerable to the security issue. After further research, Samba developers have confirmed that only Samba 3.0.2a and earlier releases contain the exploitable code.

Gael Delalleau discovered several integer overflows in Mozilla's BMP decoder that can result in denial-of-service or arbitrary code execution.

Georgi Guninski discovered a stack buffer overflow which may be triggered when viewing email messages with vCard attachments.

Several heap buffer overflows were discovered and fixed in the most recent versions of Mozilla, Firefox, and Thunderbird. These overflows may occur when:

• Using the "Send Page" function.
• Checking mail on a malicious POP3 server.
• Processing non-ASCII URLs.

Each of these vulnerabilities may be exploited for remote code execution.

Stefan Esser of e-matters discovered that PHP's strip_tags() function would ignore certain characters during parsing of tags, allowing these tags to pass through. Select browsers could then parse these tags, possibly allowing cross-site scripting attacks.

Stefan Esser of e-matters discovered a condition within PHP that may lead to remote execution of arbitrary code. The memory_limit facility is used to notify functions when memory contraints have been met. Under certain conditions, the entry into this facility is able to interrupt functions such as zend_hash_init() at locations not suitable for interruption. The result would leave these functions in a vulnerable state.

An attacker that is able to trigger the memory_limit abort within zend_hash_init() and is additionally able to control the heap before the HashTable itself is allocated, is able to supply his own HashTable destructor pointer. [...]

All mentioned places outside of the extensions are quite easy to exploit, because the memory allocation up to those places is deterministic and quite static throughout different PHP versions. [...]

Because the exploit itself consist of supplying an arbitrary destructor pointer this bug is exploitable on any platform.

In some situations, subversion metadata may be unexpectedly disclosed via WebDAV. A subversion advisory states:

mod_authz_svn, the Apache httpd module which does path-based authorization on Subversion repositories, is not correctly protecting all metadata on unreadable paths.

This security issue is not about revealing the contents of protected files: it only reveals metadata about protected areas such as paths and log messages. This may or may not be important to your organization, depending on how you're using path-based authorization, and the sensitivity of the metadata.

Source code reviews of lha by Lukasz Wojtow, Thomas Biege, and others uncovered a number of vulnerabilities affecting lha:

• Buffer overflows when handling archives and filenames. (CAN-2004-0694)
• Possible command execution via shell meta-characters when built with NOMKDIR. (CAN-2004-0745)
• Buffer overflow resulting in arbitrary code execution when handling long pathnames in LHZ archives. (CAN-2004-0769)
• Buffer overflow in the extract_one. (CAN-2004-0771)

There is a buffer overflow in the prepared statements API (libmysqlclient) when a statement containing thousands of placeholders is executed.

Under certain situations it is possible for the security icon which Mozilla displays when connected to a site using SSL to be spoofed. This could be used to make so-called "phishing attacks" more difficult to detect.

When handling FTP URLs containing NULL bytes, Mozilla will interpret the file content as HTML. This may allow unexpected execution of Javascript when viewing plain text or other file types via FTP.

A malicious web page can cause an automated file upload from the victim's machine when viewed with Mozilla with Javascript enabled. This is due to a bug permitting default values for type="file" <input> elements in certain situations.

Under some situations, Mozilla will automatically import a certificate from an email message or web site. This behavior can be used as a denial-of-service attack: if the certificate has a distinguished name (DN) identical to one of the built-in Certificate Authorities (CAs), then Mozilla will no longer be able to certify sites with certificates issued from that CA.

rssh expands command line paramters before invoking chroot. This could result in the disclosure to the client of file names outside of the chroot directory. A posting by the rssh author explains:

The cause of the problem identified by Mr. McCaw is that rssh expanded command-line arguments prior to entering the chroot jail. This bug DOES NOT allow a user to access any of the files outside the jail, but can allow them to discover what files are in a directory which is outside the jail, if their credentials on the server would normally allow them read/execute access in the specified directory.

An iDEFENSE security advisory reports:

Remote exploitation of an input validation error in version 1.2 of GNU radiusd could allow a denial of service.

The vulnerability specifically exists within the asn_decode_string() function defined in snmplib/asn1.c. When a very large unsigned number is supplied, it is possible that an integer overflow will occur in the bounds-checking code. The daemon will then attempt to reference unallocated memory, resulting in an access violation that causes the process to terminate.

A new feature of sudo 1.6.8 called "sudoedit" (a safe editing facility) may allow users to read files to which they normally have no access.

A buffer overflow exists in mod_proxy which may allow an attacker to launch local DoS attacks and possibly execute arbitrary code.

A number of vulnerabilities were discovered in CVS by Stefan Esser, Sebastian Krahmer, and Derek Price.

• Insufficient input validation while processing "Entry" lines. (CAN-2004-0414)
• A double-free resulting from erroneous state handling while processing "Argumentx" commands. (CAN-2004-0416)
• Integer overflow while processing "Max-dotdot" commands. (CAN-2004-0417)
• Erroneous handling of empty entries handled while processing "Notify" commands. (CAN-2004-0418)
• A format string bug while processing CVS wrappers.
• Single-byte buffer underflows while processing configuration files from CVSROOT.
• Various other integer overflows.

Additionally, iDEFENSE reports an undocumented command-line flag used in debugging does not perform input validation on the given path names.

CVS servers ("cvs server" or :pserver: modes) are affected by these vulnerabilities. They vary in impact but include information disclosure (the iDEFENSE-reported bug), denial-of-service (CAN-2004-0414, CAN-2004-0416, CAN-2004-0417 and other bugs), or possibly arbitrary code execution (CAN-2004-0418). In very special situations where the attacker may somehow influence the contents of CVS configuration files in CVSROOT, additional attacks may be possible.

Chris Evans discovered several flaws in the gdk-pixbuf XPM image decoder:

• Heap-based overflow in pixbuf_create_from_xpm
• Stack-based overflow in xpm_extract_color
• Integer overflows in io-ico.c

Some of these flaws are believed to be exploitable.

Chris Evans discovered several vulnerabilities in the libXpm image decoder:

• A stack-based buffer overflow in xpmParseColors
• An integer overflow in xpmParseColors
• A stack-based buffer overflow in ParsePixels and ParseAndPutPixels

The X11R6.8.1 release announcement reads:

This version is purely a security release, addressing multiple integer and stack overflows in libXpm, the X Pixmap library; all known versions of X (both XFree86 and X.Org) are affected, so all users of X are strongly encouraged to upgrade.

If the CUPS server (cupsd) receives a zero-length UDP message, it will disable its print queue browser service.

The Apache Software Foundation Security Team discovered a programming error in the apr-util library function apr_uri_parse. When parsing IPv6 literal addresses, it is possible that a length is incorrectly calculated to be negative, and this value is passed to memcpy. This may result in an exploitable vulnerability on some platforms, including FreeBSD.

A malicious user with DAV write privileges can trigger a null pointer dereference in the Apache mod_dav module. This could cause the server to become unavailable.

SITIC discovered a vulnerability in Apache 2's handling of environmental variable settings in the httpd configuration files (the main `httpd.conf' and `.htaccess' files). According to a SITIC advisory:

The buffer overflow occurs when expanding ${ENVVAR} constructs in .htaccess or httpd.conf files. The function ap_resolve_env() in server/util.c copies data from environment variables to the character array tmp with strcat(3), leading to a buffer overflow.

The Webmin developers documented a security issue in the release notes for version 1.160:

Fixed a security hole in the maketemp.pl script, used to create the /tmp/.webmin directory at install time. If an un-trusted user creates this directory before Webmin is installed, he could create in it a symbolic link pointing to a critical file on the system, which would be overwritten when Webmin writes to the link filename.

Code found in nmbd and smbd may allow a remote attacker to effectively crash the nmbd server or use the smbd server to exhaust the system memory.

zen-parse discovered a heap buffer overflow in Mozilla's POP client implementation. A malicious POP server could exploit this vulnerability to cause Mozilla to execute arbitrary code.

zen-parse discovered and iDEFENSE reported an exploitable integer overflow in a scriptable Mozilla component `SOAPParameter':

Improper input validation to the SOAPParameter object constructor in Netscape and Mozilla allows execution of arbitrary code. The SOAPParameter object's constructor contains an integer overflow which allows controllable heap corruption. A web page can be constructed to leverage this into remote execution of arbitrary code.

OpenOffice creates a working directory in /tmp on startup, and uses this directory to temporarily store document content. However, the permissions of the created directory may allow other user on the system to read these files, potentially exposing information the user likely assumed was inaccessible.

The mpg123 software version 0.59r contains a buffer overflow vulnerability which may permit the execution of arbitrary code as the owner of the mpg123 process.

Marcus Meissner discovered that ImageMagick's BMP decoder would crash when loading the test BMP file created by Chris Evans for testing the previous Qt vulnerability.

A class of bugs affecting many web browsers in the same way was discovered. A Secunia advisory reports:

The problem is that the browsers don't check if a target frame belongs to a website containing a malicious link, which therefore doesn't prevent one browser window from loading content in a named frame in another window.

Successful exploitation allows a malicious website to load arbitrary content in an arbitrary frame in another browser window owned by e.g. a trusted site.

A KDE Security Advisory reports:

A malicious website could abuse Konqueror to insert its own frames into the page of an otherwise trusted website. As a result the user may unknowingly send confidential information intended for the trusted website to the malicious website.

Secunia has provided a demonstration of the vulnerability at http://secunia.com/multiple_browsers_frame_injection_vulnerability_test/.

Numerous errors in isakmpd's input packet validation lead to denial-of-service vulnerabilities. From the Rapid7 advisory:

The ISAKMP packet processing functions in OpenBSD's isakmpd daemon contain multiple payload handling flaws that allow a remote attacker to launch a denial of service attack against the daemon.

Carefully crafted ISAKMP packets will cause the isakmpd daemon to attempt out-of-bounds reads, exhaust available memory, or loop endlessly (consuming 100% of the CPU).

Marcus Meissner discovered that imlib's BMP decoder would crash when loading the test BMP file created by Chris Evans for testing the previous Qt vulnerability. It is believed that this bug could be exploited for arbitrary code execution.

An advisory published by the MIT Kerberos team says:

The MIT Kerberos 5 implementation's Key Distribution Center (KDC) program contains a double-free vulnerability that potentially allows a remote attacker to execute arbitrary code. Compromise of a KDC host compromises the security of the entire authentication realm served by the KDC. Additionally, double-free vulnerabilities exist in MIT Kerberos 5 library code, making client programs and application servers vulnerable.

Double-free vulnerabilities of this type are not believed to be exploitable for code execution on FreeBSD systems. However, the potential for other ill effects may exist.

An advisory published by the MIT Kerberos team says:

The ASN.1 decoder library in the MIT Kerberos 5 distribution is vulnerable to a denial-of-service attack causing an infinite loop in the decoder. The KDC is vulnerable to this attack.

An unauthenticated remote attacker can cause a KDC or application server to hang inside an infinite loop.

An attacker impersonating a legitimate KDC or application server may cause a client program to hang inside an infinite loop.

Marcus Meissner discovered that imlib2's BMP decoder would crash when loading the test BMP file created by Chris Evans for testing the previous Qt vulnerability. There appears to be both a stack-based and a heap-based buffer overflow that are believed to be exploitable for arbitrary code execution.

According to the SpamAssassin 2.64 release announcement:

Security fix prevents a denial of service attack open to certain malformed messages; this DoS affects all SpamAssassin 2.5x and 2.6x versions to date.

The issue appears to be triggered by overly long message headers.

lukemftpd(8) is an enhanced BSD FTP server produced within the NetBSD project. The sources for lukemftpd are shipped with some versions of FreeBSD, however it is not built or installed by default. The build system option WANT_LUKEMFTPD must be set to build and install lukemftpd. [NOTE: An exception is FreeBSD 4.7-RELEASE, wherein lukemftpd was installed, but not enabled, by default.]

Przemyslaw Frasunek discovered several vulnerabilities in lukemftpd arising from races in the out-of-band signal handling code used to implement the ABOR command. As a result of these races, the internal state of the FTP server may be manipulated in unexpected ways.

A remote attacker may be able to cause FTP commands to be executed with the privileges of the running lukemftpd process. This may be a low-privilege `ftp' user if the `-r' command line option is specified, or it may be superuser privileges if `-r' is *not* specified.

By submitting a carefully crafted authentication packet, it is possible for an attacker to bypass password authentication in MySQL 4.1. Using a similar method, a stack buffer used in the authentication mechanism can be overflowed.

According to a Debian Security Advisory:

Andres Salomon noticed a problem in the CGI session management of Ruby, an object-oriented scripting language. CGI::Session's FileStore (and presumably PStore [...]) implementations store session information insecurely. They simply create files, ignoring permission issues. This can lead an attacker who has also shell access to the webserver to take over a session.

ISS X-Force reports that a remotely exploitable buffer overflow exists in the Netscape Security Services (NSS) library's implementation of SSLv2. From their advisory:

The NSS library contains a flaw in SSLv2 record parsing that may lead to remote compromise. When parsing the first record in an SSLv2 negotiation, the client hello message, the server fails to validate the length of a record field. As a result, it is possible for an attacker to trigger a heap-based overflow of arbitrary length.

Note that the vulnerable NSS library is also present in Mozilla-based browsers. However, it is not believed that browsers are affected, as the vulnerability is present only in code used by SSLv2 *servers*.

ripMIME may prematurely terminate decoding Base64 encoded messages when it encounters multiple blank lines or other non-standard Base64 constructs. Virus scanning and content filtering tools that use ripMIME may therefore be bypassed.

The ripMIME CHANGELOG file says:

There's viruses going around exploiting the ability to hide the majority of their data in an attachment by using blank lines and other tricks to make scanning systems prematurely terminate their base64 decoding.

The moinmoin package contains two bugs with ACLs and anonymous users. Both bugs may permit anonymous users to gain access to administrative functions; for example the delete function.

There is no known workaround, the vulnerability exists regardless if a site is using ACLs or not.

An rsync security advisory reports:

There is a path-sanitizing bug that affects daemon mode in all recent rsync versions (including 2.6.2) but only if chroot is disabled.

The bug may allow a remote user to access files outside of an rsync module's configured path with the privileges configured for that module.

Alexander Larsson reports that some versions of gnome-vfs and MidnightCommander contain a number of `extfs' scripts that do not properly validate user input. If an attacker can cause her victim to process a specially-crafted URI, arbitrary commands can be executed with the privileges of the victim.

Ulf Härnhammar discovered a pair of buffer overflows in the WAV file handling code of SoX. If an attacker can cause her victim to process a specially-crafted WAV file with SoX (e.g. through social engineering or through some other program that relies on SoX), arbitrary code can be executed with the privileges of the victim.

According to a KDE Security Advisory:

WESTPOINT internet reconnaissance services alerted the KDE security team that the KDE web browser Konqueror allows websites to set cookies for certain country specific secondary top level domains.

Web sites operating under the affected domains can set HTTP cookies in such a way that the Konqueror web browser will send them to all other web sites operating under the same domain. A malicious website can use this as part of a session fixation attack. See e.g. http://www.acros.si/papers/session_fixation.pdf

Affected are all country specific secondary top level domains that use more than 2 characters in the secondary part of the domain name and that use a secondary part other than com, net, mil, org, gov, edu or int. Examples of affected domains are .ltd.uk, .plc.uk and .firm.in

It should be noted that popular domains such as .co.uk, .co.in and .com are NOT affected.

Neils Heinen reports that the setuid `news' binaries installed as part of fidogate may be used to create files or append to file with the privileges of the `news' user by setting the LOGFILE environmental variable.

The log functions in jftpgw may allow remotely authenticated user to execute arbitrary code via the format string specifiers in certain syslog messages.

Qt contains several vulnerabilities related to image loading, including possible crashes when loading corrupt GIF, BMP, or JPEG images. Most seriously, Chris Evans reports that the BMP crash is actually due to a heap buffer overflow. It is believed that an attacker may be able to construct a BMP image that could cause a Qt-using application to execute arbitrary code when it is loaded.

An iDEFENSE security advisory describes a format string vulnerability that could be exploited when Courier-IMAP is run in debug mode (DEBUG_LOGIN set).

According to Christian Hammers:

[mysqlhotcopy created] temporary files in /tmp which had predictable filenames and such could be used for a tempfile run attack.

Jeroen van Wolffelaar is credited with discovering the issue.

Evgeny Demidov discovered that the Samba server has a buffer overflow in the Samba Web Administration Tool (SWAT) on decoding Base64 data during HTTP Basic Authentication. Versions 3.0.2 through 3.0.4 are affected.

Another buffer overflow bug has been found in the code used to support the "mangling method = hash" smb.conf option. The default setting for this parameter is "mangling method = hash2" and therefore not vulnerable. Versions between 2.2.0 through 2.2.9 and 3.0.0 through 3.0.4 are affected.

The Mozilla project's family of browsers contain a design flaw that can allow a website to spoof almost perfectly any part of the Mozilla user interface, including spoofing web sites for phishing or internal elements such as the "Master Password" dialog box. This achieved by manipulating "chrome" through remote XUL content. Recent versions of Mozilla have been fixed to not allow untrusted documents to utilize "chrome" in this way.

Chris Evans has discovered multiple vulnerabilities in libpng, which can be exploited by malicious people to compromise a vulnerable system or cause a DoS (Denial of Service).

According to a KDE Security Advisory, KDE may sometimes create temporary files without properly checking the ownership and type of the target path. This could allow a local attacker to cause KDE applications to overwrite arbitrary files.

Sebastian Krahmer discovered several remotely exploitable buffer overflow vulnerabilities in the MSN component of gaim.

In two places in the MSN protocol plugins (object.c and slp.c), strncpy was used incorrectly; the size of the array was not checked before copying to it. Both bugs affect MSN's MSNSLP protocol, which is peer-to-peer, so this could potentially be easy to exploit.

An iDEFENSE security advisory reports:

Remote exploitation of an input validation error in the uudecoding feature of Adobe Acrobat Reader (Unix) 5.0 allows an attacker to execute arbitrary code.

The Unix and Linux versions of Adobe Acrobat Reader 5.0 automatically attempt to convert uuencoded documents back into their original format. The vulnerability specifically exists in the failure of Acrobat Reader to check for the backtick shell metacharacter in the filename before executing a command with a shell. This allows a maliciously constructed filename to execute arbitrary programs.

John Graham-Cumming reports that certain configurations of POPFile may allow the retrieval of any files with the extensions .gif, .png, .ico, .css, as well as some files with the extension .html.

A buffer overflow exists in the logging functionality of the DHCP daemon which could lead to Denial of Service attacks and has the potential to allow attackers to execute arbitrary code.

Steve Grubb reports a buffer read overrun in libpng's png_format_buffer function. A specially constructed PNG image processed by an application using libpng may trigger the buffer read overrun and possibly result in an application crash.

Mozilla and Mozilla Firefox contains a flaw that may allow a malicious user to spoof SSL certification.

Glenn Randers-Pehrson has contributed a fix for the png vulnerabilities discovered by Chris Evans.

The Courier set of mail services use a common Unicode library. This library contains buffer overflows in the converters for two popular Japanese character encodings. These overflows may be remotely exploitable, triggered by a maliciously formatted email message that is later processed by one of the Courier mail services. From the release notes for the corrected versions of the Courier set of mail services:

iso2022jp.c: Converters became (upper-)compatible with ISO-2022-JP (RFC1468 / JIS X 0208:1997 Annex 2) and ISO-2022-JP-1 (RFC2237). Buffer overflow vulnerability (when Unicode character is out of BMP range) has been closed. Convert error handling was implemented.

shiftjis.c: Broken SHIFT_JIS converters has been fixed and became (upper-)compatible with Shifted Encoding Method (JIS X 0208:1997 Annex 1). Buffer overflow vulnerability (when Unicode character is out of BMP range) has been closed. Convert error handling was implemented.

Stefan Esser of e-matters Security discovered a baker's dozen of buffer overflows in Ethereal's decoders, including:

• NetFlow
• IGAP
• EIGRP
• PGM
• IRDA
• BGP
• ISUP
• TCAP
• UCP

In addition, a vulnerability in the RADIUS decoder was found by Jonathan Heusser.

Finally, there is one uncredited vulnerability described by the Ethereal team as:

A zero-length Presentation protocol selector could make Ethereal crash.

Issues have been discovered in multiple protocol dissectors.

Issues have been discovered in multiple protocol dissectors.

SSLtelnet contains a format string vulnerability that could allow remote code execution and privilege escalation.

Roman Medina-Heigl Hernandez did a survey which other webmail systems where vulnerable to a bug he discovered in SquirrelMail. This advisory summarizes the results.

When pavuk sends a request to a web server and the server sends back the HTTP status code 305 (Use Proxy), pavuk copies data from the HTTP Location header in an unsafe manner. This leads to a stack-based buffer overflow with control over EIP.

Janek Vind "waraxe" reports that several issues in the PHPNuke software may be exploited via carefully crafted URL requests. These URLs will permit the injection of SQL code, cookie theft, and the readability of the PHPNuke administrator account.

This vulnerability would allow remote user to inject PHP code to be executed by eval() function. This vulnerability is only exploitable if variable $cfg['LeftFrameLight'] is set to FALSE (in file config.inc.php).

GNATS 3.113.1 contains multiple buffer overflows, through which a local attacker could gain elevated privileges on the system.

A programming error in the handling of some Linux system calls may result in memory locations being accessed without proper validation.

It may be possible for a local attacker to read and/or overwrite portions of kernel memory, resulting in disclosure of sensitive information or potential privilege escalation. A local attacker can cause a system panic.

giFT-FastTrack is susceptible to a remote Denial of Service attack which could allow a remote attacker to render HTTP services unusable. According to the developers, no code execution is possible; however, they recommend an immediate upgrade.

When the IPv6 code was added to xdm a critical test to disable xdmcp was accidentally removed. This caused xdm to create the chooser socket regardless if DisplayManager.requestPort was disabled in xdm-config or not.

A serious flaw exists in the MoinMoin software which may allow a malicious user to gain access to unauthorized privileges.

In December 2002, Timo Sirainen reported:

Cyrus IMAP server has a a remotely exploitable pre-login buffer overflow. [...] Note that you don't have to log in before exploiting this, and since Cyrus runs everything under one UID, it's possible to read every user's mail in the system.

It is unknown whether this vulnerability is exploitable for code execution on FreeBSD systems.

The Cyrus team reported multiple vulnerabilities in older versions of Cyrus IMSPd:

These releases correct a recently discovered buffer overflow vulnerability, as well as clean up a significant amount of buffer handling throughout the code.

A remotely exploitable heap buffer overflow vulnerability was found in MPlayer's URL decoding code. If an attacker can cause MPlayer to visit a specially crafted URL, arbitrary code execution with the privileges of the user running MPlayer may occur. A `visit' might be caused by social engineering, or a malicious web server could use HTTP redirects which MPlayer would then process.

Timo Sirainen reports multiple buffer overflows that may be triggered while parsing messages, as well as input validation errors that could result in disclosure of mailing list passwords.

These bugs were resolved in the August 2003 snapshot of ecartis.

When the directive "SecFilterScanPost" is enabled, the Apache 2.x version of ModSecurity is vulnerable to an off-by-one overflow

clamav will exit when a programming assertion is not met. A malformed uuencoded message can trigger this assertion, allowing an attacker to trivially crash clamd or other components of clamav.

Stefan Esser reports:

A vulnerability within a libneon date parsing function could cause a heap overflow which could lead to remote code execution, depending on the application using libneon.

The vulnerability is in the function ne_rfc1036_parse, which is in turn used by the function ne_httpdate_parse. Applications using either of these neon functions may be vulnerable.

Greuff reports that the neon WebDAV client library contains several format string bugs within error reporting code. A malicious server may exploit these bugs by sending specially crafted PROPFIND or PROPPATCH responses.

Although several applications include neon, such as cadaver and subversion, the FreeBSD Ports of these applications are not impacted. They are specifically configured to NOT use the included neon. Only packages listed as affected in this notice are believed to be impacted.

A flaw exists in Gallery versions previous to 1.4.3-pl1 and post 1.2 which may give an attacker the potential to log in under the "admin" account. Data outside of the gallery is unaffected and the attacker cannot modify any data other than the photos or photo albums.

Jakub Jelinek reports several security related bugs in Midnight Commander, including:

• Multiple buffer overflows (CAN-2004-0226)
• Insecure temporary file handling (CAN-2004-0231)
• Format string bug (CAN-2004-0232)

Remote exploitation of a buffer overflow vulnerability in the NTLM authentication helper routine of the Squid Web Proxy Cache could allow a remote attacker to execute arbitrary code. A remote attacker can compromise a target system if the Squid Proxy is configured to use the NTLM authentication helper. The attacker can send an overly long password to overflow the buffer and execute arbitrary code.

The NISCC and the OUSPG developed a test suite for the H.323 protocol. This test suite has uncovered vulnerabilities in several H.323 implementations with impacts ranging from denial-of-service to arbitrary code execution.

In the FreeBSD Ports Collection, `pwlib' is directly affected. Other applications such as `asterisk' and `openh323' incorporate `pwlib' statically and so are also independently affected.

A programming error resulting in a failure to verify that an attempt to manipulate routing tables originated from a non-jailed process.

Jailed processes running with superuser privileges could modify host routing tables. This could result in a variety of consequences including packets being sent via an incorrect network interface and packets being discarded entirely.

Programming errors in the implementation of the msync(2) system call involving the MS_INVALIDATE operation lead to cache consistency problems between the virtual memory system and on-disk contents.

In some situations, a user with read access to a file may be able to prevent changes to that file from being committed to disk.

The leafnode NNTP server may go into an unterminated loop with 100% CPU use when an article is requested by Message-ID that has been crossposted to several news groups when one of the group names is the prefix of another group name that the article was cross-posted to. Found by Jan Knutar.

Fetchnews could hang when a news article to be downloaded lacked one of the mandatory headers. Found by Joshua Crawford.

When a downloaded news article ends prematurely, i. e. when the server sends [CR]LF.[CR]LF before sending a blank line, fetchnews may wait indefinitely for data that never arrives. Workaround: configure "minlines=1" (or use a bigger value) in the configuration file. Found by Toni Viemerö.

Shaun Colley reports that the script `mysqlbug' included with MySQL sometimes creates temporary files in an unsafe manner. As a result, an attacker may create a symlink in /tmp so that if another user invokes `mysqlbug' and quits without making any changes, an arbitrary file may be overwritten with the bug report template.

Stefan Esser reports:

Subversion versions up to 1.0.2 are vulnerable to a date parsing vulnerability which can be abused to allow remote code execution on Subversion servers and therefore could lead to a repository compromise.

NOTE: This vulnerability is similar to the date parsing issue that affected neon. However, it is a different and distinct bug.

Due to a programming error in code used to parse data received from the client, malformed data can cause a heap buffer to overflow, allowing the client to overwrite arbitrary portions of the server's memory.

A malicious CVS client can exploit this to run arbitrary code on the server at the privilege level of the CVS server software.

Joe Orton reports a memory leak in Apache 2's mod_ssl. A remote attacker may issue HTTP requests on an HTTPS port, causing an error. Due to a bug in processing this condition, memory associated with the connection is not freed. Repeated requests can result in consuming all available memory resources, probably resulting in termination of the Apache process.

Karol Wiesek and Greg MacManus reported via iDEFENSE that the Opera web browser contains a flaw in the handling of certain URIs. When presented with these URIs, Opera would invoke external commands to process them after some validation. However, if the hostname component of a URI begins with a `-', it may be treated as an option by an external command. This could have undesirable side-effects, from denial-of-service to code execution. The impact is very dependent on local configuration.

After the iDEFENSE advisory was published, the KDE team discovered similar problems in KDE's URI handlers.

The Debian security team reported a pair of vulnerabilities in fsp:

A vulnerability was discovered in fsp, client utilities for File Service Protocol (FSP), whereby a remote user could both escape from the FSP root directory (CAN-2003-1022), and also overflow a fixed-length buffer to execute arbitrary code (CAN-2004-0011).

Jindrich Makovicka reports a regression in proftpd's handling of IP address access control lists (IP ACLs). Due to this regression, some IP ACLs are treated as ``allow all''.

Some scripts installed with xine create temporary files insecurely. It is recommended that these scripts (xine-check, xine-bugreport) not be used. They are not needed for normal operation.

A remote exploitable buffer overflow has been discovered in exim when verify = header_syntax is used in the configuration file. This does not affect the default configuration.

The includes/sessions.php unnecessarily adds session item into session table and therefore vulnerable to a denial-of-service attack.

An input validation error was discovered in the kadmind code that handles the framing of Kerberos 4 compatibility administration requests. The code assumed that the length given in the framing was always two or more bytes. Smaller lengths will cause kadmind to read an arbitrary amount of data into a minimally-sized buffer on the heap.

A remote attacker may send a specially formatted message to kadmind, causing it to crash or possibly resulting in arbitrary code execution.

The kadmind daemon is part of Kerberos 5 support. However, this bug will only be present if kadmind was built with additional Kerberos 4 support. Thus, only systems that have *both* Heimdal Kerberos 5 and Kerberos 4 installed might be affected.

NOTE: On FreeBSD 4 systems, `kadmind' may be installed as `k5admind'.

Two programming errors were discovered in which path names handled by CVS were not properly validated. In one case, the CVS client accepts absolute path names from the server when determining which files to update. In another case, the CVS server accepts relative path names from the client when determining which files to transmit, including those containing references to parent directories (`../').

These programming errors generally only have a security impact when dealing with remote CVS repositories.

A malicious CVS server may cause a CVS client to overwrite arbitrary files on the client's system.

A CVS client may request RCS files from a remote system other than those in the repository specified by $CVSROOT. These RCS files need not be part of any CVS repository themselves.

The kernel interface for creating a snapshot of a filesystem is the same as that for changing the flags on that filesystem. Due to an oversight, the mksnap_ffs(8) command called that interface with only the snapshot flag set, causing all other flags to be reset to the default value.

A regularly scheduled backup of a live filesystem, or any other process that uses the mksnap_ffs command (for instance, to provide a rough undelete functionality on a file server), will clear any flags in effect on the filesystem being snapshot. Possible consequences depend on local usage, but can include disabling extended access control lists or enabling the use of setuid executables stored on an untrusted filesystem.

The mksnap_ffs command is normally only available to the superuser and members of the `operator' group. There is therefore no risk of a user gaining elevated privileges directly through use of the mksnap_ffs command unless it has been intentionally made available to unprivileged users.

A programming error in the shmat(2) system call can result in a shared memory segment's reference count being erroneously incremented.

It may be possible to cause a shared memory segment to reference unallocated kernel memory, but remain valid. This could allow a local attacker to gain read or write access to a portion of kernel memory, resulting in sensitive information disclosure, bypass of access control mechanisms, or privilege escalation.

A programming error has been found in the jail_attach(2) system call which affects the way that system call verifies the privilege level of the calling process. Instead of failing immediately if the calling process was already jailed, the jail_attach system call would fail only after changing the calling process's root directory.

A process with superuser privileges inside a jail could change its root directory to that of a different jail, and thus gain full read and write access to files and directories within the target jail.

FreeBSD does not limit the number of TCP segments that may be held in a reassembly queue. A remote attacker may conduct a low-bandwidth denial-of-service attack against a machine providing services based on TCP (there are many such services, including HTTP, SMTP, and FTP). By sending many out-of-sequence TCP segments, the attacker can cause the target machine to consume all available memory buffers (``mbufs''), likely leading to a system crash.

From the FreeBSD Security Advisory:

A programming error in the handling of some IPv6 socket options within the setsockopt(2) system call may result in memory locations being accessed without proper validation.

It may be possible for a local attacker to read portions of kernel memory, resulting in disclosure of sensitive information. A local attacker can cause a system panic.

A remote attacker could cause an application using OpenSSL to crash by performing a specially crafted SSL/TLS handshake.

A programming error in BIND 8 named can result in a DNS message being incorrectly cached as a negative response. As a result, an attacker may arrange for malicious DNS messages to be delivered to a target name server, and cause that name server to cache a negative response for some target domain name. The name server would thereafter respond negatively to legitimate queries for that domain name, resulting in a denial-of-service for applications that require DNS.

Heimdal does not correctly validate the `transited' field of Kerberos tickets when computing the authentication path. This could allow a rogue KDC with which cross-realm relationships have been established to impersonate any KDC in the authentication path.

Ulf Härnhammar discovered several vulnerabilities in LHa for UNIX's path name handling code. Specially constructed archive files may cause LHa to overwrite files or execute arbitrary code with the privileges of the user invoking LHa. This could be particularly harmful for automated systems that might handle archives such as virus scanning processes.

A straightforward stack buffer overflow exists in XChat's Socks5 proxy support.

The XChat developers report that `tsifra' discovered this issue.

NOTE: XChat Socks5 support is disabled by support in the FreeBSD Ports Collection.

When running rsync in daemon mode, no checks were made to prevent clients from writing outside of a module's `path' setting.

From the xinehq advisory:

By opening a malicious MRL in any xine-lib based media player, an attacker can write arbitrary content to an arbitrary file, only restricted by the permissions of the user running the application.

The flaw is a result of a feature that allows MRLs (media resource locator URIs) to specify arbitrary configuration options.

An unknown remotely exploitable vulnerability was disclosed. Robert Segall writes:

a security vulnerability was brought to my attention (many thanks to Akira Higuchi). Everyone running any previous version should upgrade to 1.6 immediately - the vulnerability may allow a remote exploit. No exploits are currently known and none have been observed in the wild till now. The danger is minimised if you run Pound in a root jail and/or you run Pound as non-root user.

The common.php script always trusts the `X-Forwarded-For' header in the client's HTTP request. A remote user could forge this header in order to bypass any IP address access control lists (ACLs).

NISCC / UNIRAS has published an advisory that re-visits the long discussed spoofed TCP RST denial-of-service vulnerability. This new look emphasizes the fact that for some applications such attacks are practically feasible.

Jack of RaptureSecurity reported a double byte buffer overflow in ident2. The bug may allow a remote attacker to execute arbitrary code within the context of the ident2 daemon. The daemon typically runs as user-ID `nobody', but with group-ID `wheel'.

A buffer overflow is present in some versions of the KDE personal information manager (kdepim) which may be triggered when processing a specially crafted VCF file.

When racoon receives an ISAKMP header, it will attempt to allocate sufficient memory for the entire ISAKMP message according to the header's length field. If an attacker crafts an ISAKMP header with a ridiculously large value in the length field, racoon may exceed operating system resource limits and be terminated, resulting in a denial of service.

When racoon receives an IKE message with an incorrectly constructed Generic Payload Header, it may behave erratically, going into a tight loop and dropping connections.

Chad Loder has discovered vulnerabilities in tcpdump's ISAKMP protocol handler. During an audit to repair these issues, Bill Fenner discovered some related problems.

These vulnerabilities may be used by an attacker to crash a running `tcpdump' process. They can only be triggered if the `-v' command line option is being used.

NOTE: the racoon ISAKMP/IKE daemon incorporates the ISAKMP protocol handler from tcpdump, and so is also affected by this issue.

Midnight Commander uses a fixed sized stack buffer while resolving symbolic links within file archives (tar or cpio). If an attacker can cause a user to process a specially crafted file archive with Midnight Commander, the attacker may be able to obtain the privileges of the target user.

Ralf Spenneberg discovered a serious flaw in racoon. When using Phase 1 main or aggressive mode, racoon does not verify the client's RSA signature. Any installations using X.509 authentication are strongly urged to upgrade.