summaryrefslogtreecommitdiff
diff options
context:
space:
mode:
authorKris Kennaway <kris@FreeBSD.org>2000-01-11 07:46:33 +0000
committerKris Kennaway <kris@FreeBSD.org>2000-01-11 07:46:33 +0000
commit3aa5f62f030917e219638b2bcf18e3c58532556c (patch)
tree2d165a51ddf99c1ae8200014db3c80e9a391c65a
parentfac8edac5e1d7ef65495007fea4ff3e72184cac5 (diff)
Notes
-rw-r--r--usr.sbin/ctm/ctm/ctm.128
-rw-r--r--usr.sbin/ctm/ctm_rmail/ctm_rmail.147
2 files changed, 51 insertions, 24 deletions
diff --git a/usr.sbin/ctm/ctm/ctm.1 b/usr.sbin/ctm/ctm/ctm.1
index 4656301c11ba..d51bbc22bf4c 100644
--- a/usr.sbin/ctm/ctm/ctm.1
+++ b/usr.sbin/ctm/ctm/ctm.1
@@ -222,7 +222,33 @@ Pathnames can be selected for CTM's consideration using the
option.
.El
-
+.Pp
+.Sh SECURITY
+.Pp
+CTM is an
+.Bf Em
+INSECURE PROTOCOL
+.Ef
+- there is no authentication performed that the
+changes applied to the source code were sent by a
+trusted party, and so care should be taken if the
+CTM deltas are obtained via an unauthenticated
+medium such as email.
+It is a relatively simple matter for an attacker
+to forge a CTM delta to replace or precede the
+legitimate one and insert malicious code into your
+source tree.
+If the legitimate delta is somehow prevented from
+arriving, this will go unnoticed until a later
+delta attempts to touch the same file, at which
+point the MD5 checksum will fail.
+.Pp
+A future version of
+.Fx
+may solve this problem by authenticating CTM
+deltas using cryptographic signatures, but in the
+mean time it is strongly recommended that you
+obtain the CTM deltas via FTP, and not via email.
.Sh ENVIRONMENT
.Ev TMPDIR,
if set to a pathname, will cause ctm to use that pathname
diff --git a/usr.sbin/ctm/ctm_rmail/ctm_rmail.1 b/usr.sbin/ctm/ctm_rmail/ctm_rmail.1
index 77a670baab19..49d60a664be7 100644
--- a/usr.sbin/ctm/ctm_rmail/ctm_rmail.1
+++ b/usr.sbin/ctm/ctm_rmail/ctm_rmail.1
@@ -364,30 +364,31 @@ to execute
.Xr ctm
on the (non-FreeBSD) machine that this example was taken from.
.Sh SECURITY
-If you automatically take your mail and pass it to a file tree patcher, you
-might think you are handing the keys to your system to the crackers! Happily,
-the window for mischief is quite small.
-.Nm ctm_rmail
-is careful to write only to the directories given to it (by not believing any
-.Dq /
-characters in the delta name), and the latest
-.Xr ctm
-disallows absolute pathnames and
-.Dq \&\.\.
-in files it manipulates, so the worst you
-could lose are a few source tree files (recoverable from your deltas).
-Since
-.Xr ctm
-requires that a
-.Xr md5
-checksum match before it touches a file, only fellow
-source recipients would be able to generate a fake delta, and they're such
-nice folk that they wouldn't even think of it! :-)
.Pp
-Even this possibility could be removed by using cryptographic signatures.
-A possible future enhancement would be to use
-.Nm PGP
-to provide a secure wrapper.
+CTM is an
+.Bf Em
+INSECURE PROTOCOL
+.Ef
+- there is no authentication performed that the
+changes applied to the source code were sent by a
+trusted party, and so care should be taken if the
+CTM deltas are obtained via an unauthenticated
+medium such as email.
+It is a relatively simple matter for an attacker
+to forge a CTM delta to replace or precede the
+legitimate one and insert malicious code into your
+source tree.
+If the legitimate delta is somehow prevented from
+arriving, this will go unnoticed until a later
+delta attempts to touch the same file, at which
+point the MD5 checksum will fail.
+.Pp
+A future version of
+.Fx
+may solve this problem by authenticating CTM
+deltas using cryptographic signatures, but in the
+mean time it is strongly recommended that you
+obtain the CTM deltas via FTP, and not via email.
.\" This next request is for sections 1, 6, 7 & 8 only
.Sh ENVIRONMENT
If deltas are to be applied then