[lkml]   [1997]   [Feb]   [10]   [last100]   RSS Feed
Views: [wrap][no wrap]   [headers]  [forward] 
Messages in this thread
SubjectRe: Bliss Virus (fwd)

hi all

> > errors. Can we get CERT (LERT?) to try this "virus" code,
> > so that if it turns out to be legitimate, proper warnings can
> > be disseminated through the proper (credible) channels?

it's more of a trojan than a virus, but anyway...

> > If asked, will the poster of this message provide virus code
> > to the members of this list? I expect this virus to suddenly,
> > mysteriously disappear, based on that question.

i couldn;t find any source, therefore this analysis is based on the
uuencoded binary posted here the other day. the exact details are still a
bit hazy, but here's the general idea:


this virus is a complete ELF executable with voluminous debugging
comments. it looks for executables you have write access to,
prepends itself to them, and appends a signature. when run, it
extracts the original executable into a /tmp/.bliss-tmp.<pid> file
and forks. the parent execs the temp file while the child keeps
running the infected original.

it does a bit of stealth if you try process tracing, and looks like
it wants to patch umask() in /usr/src/linux/kernel/sys.c. it does some
.rhosts stuff for network propagation. it scans your path for
executables, plus it loops over the getpwent() db looking for bin
directories and .rhosts files.

looks like a new virus infecting an older file will upgrade the old
file, and an old virus infecting a newer file will upgrade itself
from the newer file.


it seems that this version is a bit different from the one described on
the mcafee site. the size is different (18604 bytes) and it claims to be:

bliss type 1 version 0.4.0 (00010004)
Compiled on Feb 5 1997 at 18:35:33

the mcafee site says the virus destructively overwrites the
executable. this version prepends itself to executables and appends a
signature to the file stating the version that the file is infected
with. the uninfect option checks the signature, then strips the
virus and the signature to restore the original executable.

when you run an infected executable, it forks. the original exec is
extracted to /tmp/.bliss-tmp.<pid> and executed, while the child keeps
running the infected file.

infected ELF execs seem to run okay, but infected shell scripts
complain about the appended signature when you run them. i guess the
ELF loader reads only the sections in its 'table of contents' but the
#! loader just dumps the rest of the file to the interpreter?

if you trace the process (task->flags & 0x30) the virus exits. change

00002ae0: 740e b801 0000 00eb 098d b426 0000 0000

to this:

00002ae0: 740e b800 0000 00eb 098d b426 0000 0000

to disable the check.

it keeps an infection log in (by default) /tmp/.bliss which uninfect uses
to clean up infected files. each record contains the infection time,
bliss version, and pathname of the infected file.

command line args start with --bliss-, equivalent args are listed
separated by commas

uninfect-files-please, disinfect-files-please
dont-run-original, just-run-bliss, just-run-virus
dont-run-virus, dont-run-bliss, just-run-original
infect-file <file-list>

unrecognized args containing please will exit(0), otherwise the virus
will run

there is some (seemingly unfinished??) code in place to patch
sys_umask() in /usr/src/linux/kernel/sys.c along these lines:

"if(mask&023000) {current->uid = current->euid = current->suid =
current->fsuid = 0; return old&023000;}"

the worm code uses rsh and ssh. it checks /etc/hosts.equiv,
$HOME/.rhosts and scans getpwent() checking for a .rhosts in all home
directories. haven't examined this in detail yet.

finally, it checks

each getpwent()/bin

looking for executable files to infect. it seems to hit anything
executable that you can write to. this version will scan 20
directories and then stop.


Ray Lehtiniemi <>

 \ /
  Last update: 2005-03-22 13:38    [W:0.043 / U:4.196 seconds]
©2003-2020 Jasper Spaans|hosted at Digital Ocean and TransIP|Read the blog|Advertise on this site