Sysop: | Amessyroom |
---|---|
Location: | Fayetteville, NC |
Users: | 28 |
Nodes: | 6 (0 / 6) |
Uptime: | 55:31:44 |
Calls: | 422 |
Files: | 1,025 |
Messages: | 90,708 |
Package: libapache-sessionx-perl
Version: 2.01-5
Severity: important
Tags: security
Hi,
As discussed in oss-security[1], libapache-sessionx-perl uses a poor
source of entropy in Apache::Session::Generate::MD5. The critical part
is moving away from rand (e.g. to using urandom), but it would also be
a good time to update the way the id is generated.
The details are in the oss-sec thread.
[1] https://www.openwall.com/lists/oss-security/2019/06/15/1
On Mon, Jun 17, 2019 at 10:44:52PM +0200, Raphael Geissert wrote:
Package: libapache-sessionx-perl
Version: 2.01-5
Severity: important
Tags: security
Hi,
As discussed in oss-security[1], libapache-sessionx-perl uses a poor
source of entropy in Apache::Session::Generate::MD5. The critical part
is moving away from rand (e.g. to using urandom), but it would also be
a good time to update the way the id is generated.
The details are in the oss-sec thread.
[1] https://www.openwall.com/lists/oss-security/2019/06/15/1
AFAICS libapache-sessionx-perl only exists to support libembperl-perl.
As we're not going to ship libembperl-perl in trixie due to #1042845,
I wonder if we should remove libapache-sessionx-perl from testing too?
Alternatively, the approach taken for libapache-session-perl #930659
(using Crypt::URandom) seems easy to apply here as well.
https://sources.debian.org/src/libapache-session-perl/1.94-2/debian/patches/use-crypt-urandom.patch/