[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]
Message-ID: <442A2113.3070908@album.co.nz>
Date: Wed Mar 29 06:54:37 2006
From: jasper at album.co.nz (Jasper Bryant-Greene)
Subject: Critical PHP bug - act ASAP if you are running
web with sensitive data
T?nu Samuel wrote:
> Jasper Bryant-Greene wrote:
>
>> My point is, can you think of a logical reason why html_entity_decode
>> would be run on user input? I'm sure some idiot is doing it (and
>> therefore this is a security issue, though not exactly critical), but
>> I don't think I can think of a reason why it would be done.
>>
>> Why would you want to decode HTML entities given by a user? The
>> opposite (encode their input into HTML entities) is the usual approach...
>
> Ok, this "critical" is my fault. Seeing memory dump of other user data
> seems serious enough to me and I suspected it might affect different
> functions despite this one. Now when we know more, I agree that it is
> less critical than suspected by me. Still it is a problem and as subject
> told: "if you are running web with sensitive data". Malicious user can
> upload new script and see what others are doing. In most cases not so
> critical as I assumed but still bad enough and I really expect to see
> announcements for such problems faster and patches to come out (I mean
> RPM-s this time). Right now my systems are unprotected till I start to
> make packages myself or Novell is going to make one. Three weeks is too
> much. And what about PHP 4.x and 5.0 users?
Sure, this is still a fairly serious bug. (As an aside, if you have
sensitive data, you really shouldn't allow users to upload new scripts,
or be running in a shared hosting env.)
I can't speak for other distros, but there's a bug in Gentoo Bugzilla
for this: http://bugs.gentoo.org/127939
Jasper
Powered by blists - more mailing lists