Testing zlib vulnerability

From: Mathieu Lafon (Mathieu.Lafonat_private)
Date: Thu Mar 21 2002 - 12:38:19 PST

  • Next message: Steven M. Christey: "Re: useless security@ contacts"

    I use a proprietary virus scanner that I higly suspect to statically use
    zlib 1.1.3 (according to find-zlib.pl). The vendor tells me 'he thinks'
    it is not vulnerable but i want to be sure by myself (and to force him to
    release a version compiled with zlib 1.1.4).
    
    I tried to create an evil gz file by using a standard gz header, followed
    by random data and test it against minigzip (included in zlib sources).
    
    Surprisingly, it is very easy to create a file that crash zlib :
    
    ( cat *.c | gzip -9 | dd bs=1 count=80 ; dd if=/dev/urandom bs=1 count=100 ) | ./minigzip -d
    Segmentation fault
    
    With count=80, it's only a few tries.
    
    I also try this with zlib 1.1.4 and (hopefuly) only got
    ./minigzip: failed gzclose
    
    Once i got an evil gz file, i tried my closed-source virus scanner but i have
    not been able to crash it.
    
    Is there any reason that my scanner does not crash if it uses zlib-1.1.3 ?
    
    Does someone already try (and succeed) to crash a program this way instead of
    trying to detect zlib in the binary ?
    
    Thanks,
    -- 
    Mathieu Lafon
    



    This archive was generated by hypermail 2b30 : Thu Mar 21 2002 - 16:11:01 PST