Fix potential PNG decompression DOS #1060
Conversation
|
Thanks @wiredfool. Re: 1, Right now I'm more concerned about preventing a DOS than providing a consistent API and re: 2, can we reasonably defend against images with many-just-under-the-limit text segments? |
|
That was less invasive than I'd thought. Implementation allowing 64 * maxsize total memory usage for text chunks, and a test attempting to save 128 of them. ( ~8k or so of compressed text chunks in the image would expand to ~128M) |
|
|
| im = Image.open(test_file) | ||
| im.load() | ||
| except ValueError as msg: | ||
| self.assert_(msg, "Decompressed Data Too Large") |
There was a problem hiding this comment.
assertTrue is a bit clearer (for me) than assert_ here.
|
👍 in general, with minor review notes. Did you want |
|
They're not especially slow, and as most of the DOS ones go, it's not a problem once we've solved the issue. They may be a bit more of an issue running the tests in a resource constrained environment, so I'd rather not run them by default. |
|
Changes Unknown when pulling 6696b78 on wiredfool:png-dos into * on python-pillow:master*. |
Fix potential PNG decompression DOS
|
Note that this has had a CVE assigned: CVE-2014-9601 (http://www.cvedetails.com/cve/CVE-2014-9601/) |
@hugovk @aclark4life
Two outstanding bits:
Opinions?
(note that I had to change the test to pass -- I believe that the new version is also a reasonable representation of an invalid iTXt segment)