An out of bounds memory access issue issue exists in WebKit's handling of HTML tables. Visiting a maliciously crafted website may lead to an unexpected application termination or arbitrary code execution. This issue is addressed through improved bounds checking. References: Bugzilla: https://bugs.webkit.org/show_bug.cgi?id=38261 Trac: http://trac.webkit.org/changeset/59495 Acknowledgements: Red Hat would like to thank Drew Yao of Apple Product Security for responsibly reporting this issue. Upstream acknowledges wushi of team509 as the original reporter.
This is likely a crash only, but it seems it doesn't affect RHEL5 or RHEL6. This snipped appears to save the day: if( cCol + i >= nEffCols ) { table->appendColumn( span - usedSpan ); nEffCols++; width.resize( nEffCols ); width[nEffCols-1] = Length(); } As the value of i increases, so does the size of width. This should mean that the worst outcome of this flaw would be an OOB read, not an OOB write. An OOB read is a DoS only and not security for end user applications. It someone sees a flaw in my analysis, please speak up.
The same code above exists in KDE4's KHTML as well (khtml/rendering/table_layout.cpp) so would not affect KDE3 or KDE4.
Public via: [1] http://support.apple.com/kb/HT4196