- From: Julian Reschke <julian.reschke@gmx.de>
- Date: Sat, 26 Nov 2005 10:14:17 +0100
- To: Cullen Jennings <fluffy@cisco.com>
- CC: bugzilla@soe.ucsc.edu, WebDav <w3c-dist-auth@w3.org>
Cullen Jennings wrote: > > Do you think it is wrong? It's not necessarily wrong, but it definitively needs to be discussed and understood. This is a Security Consideration, and if we add something new *here*, we better make sure it's correct. Quoting: <http://greenbytes.de/tech/webdav/draft-ietf-webdav-rfc2518bis-08.html#rfc.section.19.8>: > HTTP has the ability to host scripts which are executed on client > machines. These scripts can be used to read another user's cookies > and therefore may provide an attacker the ability to use another > user's session, assume their identity temporarily and gain access to > their resources. Other attacks are also possible with client- > executed scripts. 1) How would a script be able to read another user's cookies? If this is about potential bugs in user agents, it should say so. The way it's written now confuses me, because I don't understand the risk yet. 2) And yes, letting people store arbitrary content on a shared server introduces the same security problems when people share a file server, or people download code from the Web. In particular, this includes executables, or file types with security issues (such as the recent JPG-related attacks). RFC2518bis *could* talk about that, but then this really needs to be expanded a lot. > WebDAV may worsen this situation only by making it easier for a Web > server to host content provided by many different authors (making it > harder to trust the content providers) and to host content with > restricted access alongside public pages (see particularly RFC3744). How is the latter a security problem? > HTTP servers may mitigate some of these threats by filtering content > in areas where many authors contribute pages -- the server could, for > example, remove script from HTML pages. It could also run virus scanners on new content. > This vulnerability should provide yet another reason for server > implementors and administrators not to replace authentication > mechanisms with cookie-based session tokens if there's any sensitive > information relying on the authenticated identity. This relates back to the cookie discussion above. I prefer HTTP auth as well, but I don't think this is the right place to evangelize it. > HTTP and WebDAV client implementors might consider locking down the > use of scripts and cookies based on these considerations. Best regards, Julian
Received on Saturday, 26 November 2005 09:15:17 UTC