Concurrency-related vulnerabilities in browsers - expect problems

2006.08.23
Risk: High
Local: No
Remote: Yes
CWE: CWE-264


CVSS Base Score: 7.6/10
Impact Subscore: 10/10
Exploitability Subscore: 4.9/10
Exploit range: Remote
Attack complexity: High
Authentication: No required
Confidentiality impact: Complete
Integrity impact: Complete
Availability impact: Complete

Good morning, "Fame-hungry sociopath torches cars, finds browser flaws WARSAW, Poland (AP) -- police are on a look out for a local adolescent vandal who continues to terrorize local IT workers in what appears to be a bizzare bid for fame. Larry Seltzer reports from the scene." Well, I just had to do this, forgive me. There seems to be an interesting class of concurrency-related bugs in popular browsers. This is quite similar to signal-handling flaws you might be familiar with: many browser events can be triggered asynchronously, for example using Javascript timers, while some components of the browser are still running. In many cases, a new action might be initiated that interferes with or counters the interrupted (or still executing) task. Problems like this may leave the program in inconsistent state, and later cause double frees or related issues. That usually opens the door to system compromise through careful manipulation of memory contents. The attacks would depend heavily on network latency and jitter, but can be executed. Given that the tip of that iceberg has been probed recently - for example here: http://www.mozilla.org/security/announce/2006/mfsa2006-46.html - I assumed it is now the time to post my older example. A fairly reliable example is when Firefox is interrupted by a Javascript handler while parsing a deeply nested XML document for display. If the browser is then redirected from the script to a new location, the unfinished parsing process is aborted, and all its structures are freed - but these were not left in the expected state by the parser. This is a demo that will usually crash Firefox in a couple of seconds (SEGV on Linux and MacOS, silent crashes on Windows): http://lcamtuf.coredump.cx/ffoxdie.html Have fun! PS. For the easily amused: MSIE loves "<DT><H1 STYLE=width:1px><LI></H1>" /mz http://lcamtuf.coredump.cx/ Update : Here's another separate issue that typically causes fault on memory access to website-influenced memory access: http://lcamtuf.coredump.cx/ffoxdie3.html This is separate from the previously presented example (which, remarkably, also had a tendency to trigger an unrelated call stack overflow due to XML parsing glitch on some platforms, which caused some confusion - my bad). Note that because it depends on timing more heavily, it may not work in the first shot on all computers (though it should). /mz //SecurityReason Note : All Exploits included in ExploitAlert


Vote for this issue:
50%
50%


 

Thanks for you vote!


 

Thanks for you comment!
Your message is in quarantine 48 hours.

Comment it here.


(*) - required fields.  
{{ x.nick }} | Date: {{ x.ux * 1000 | date:'yyyy-MM-dd' }} {{ x.ux * 1000 | date:'HH:mm' }} CET+1
{{ x.comment }}

Copyright 2024, cxsecurity.com

 

Back to Top