Format string vulnerability in iMovie HD 6.0.3, and Safari in Apple Mac OS X 10.4 through 10.4.10, allows remote user-assisted attackers to cause a denial of service (crash) via format string specifiers in a filename, which is not properly handled when calling the NSRunCriticalAlertPanel Apple AppKit function.
WebCore on Apple Mac OS X 10.3.9 and 10.4.10, as used in Safari, does not properly parse HTML comments in TITLE elements, which allows remote attackers to conduct cross-site scripting (XSS) attacks and bypass some XSS protection schemes by embedding certain HTML tags within an HTML comment.
Integer overflow in ImageIO in Apple Mac OS X 10.4 up to 10.4.5 allows remote attackers to cause a denial of service (crash) via a crafted JPEG image with malformed JPEG metadata, as demonstrated using Safari, aka "Deja-Doom".
The khtml::RenderTableSection::ensureRows function in KHTMLParser in Apple Mac OS X 10.4.3 and earlier, as used by Safari and TextEdit, allows remote attackers to cause a denial of service (memory consumption and application crash) via HTML files with a large ROWSPAN attribute in a TD tag.
Safari in Mac OS X 10.3.9 and 10.4.2, when rendering Rich Text Format (RTF) files, can directly access URLs without performing the normal security checks, which allows remote attackers to execute arbitrary commands.
Safari in Mac OS X 10.3.9 and 10.4.2 submits forms from an XSL formatted page to the next page that is browsed by the user, which causes form data to be sent to the wrong site.
Safari in WebKit in Mac OS X 10.4 to 10.4.2 directly accesses URLs within PDF files without the normal security checks, which allows remote attackers to execute arbitrary code via links in a PDF file.
Safari 1.2.4 on Mac OS X 10.3.6 allows remote attackers to cause a denial of service (application crash from memory exhaustion), as demonstrated using Javascript code that continuously creates nested arrays and then sorts the newly created arrays.
Safari 1.x allows remote attackers to spoof arbitrary web sites by injecting content from one window into a target window whose name is known but resides in a different domain, as demonstrated using a pop-up window on a trusted web site, aka the "window injection" vulnerability, a different vulnerability than CVE-2004-1122.