WebCore, as used in Apple Safari before 3.1, does not properly mask the password field when reverse conversion is used with the Kotoeri input method, which allows physically proximate attackers to read the password.

WebCore, as used in Apple Safari before 3.1, does not enforce the frame navigation policy for Java applets, which allows remote attackers to conduct cross-site scripting (XSS) attacks.

Cross-site scripting (XSS) vulnerability in WebCore, as used in Apple Safari before 3.1, allows remote attackers to inject arbitrary web script or HTML via unknown vectors related to sites that set the document.domain property or have the same document.domain.

Buffer overflow in WebKit, as used in Apple Safari before 3.1, allows remote attackers to execute arbitrary code via crafted regular expressions in JavaScript.

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.

Apple Safari allows remote attackers to cause a denial of service (application crash) via a crafted data:// URL.

AppleWebKit (WebCore and WebKit), as used in multiple products such as Safari 1.2 and OmniGroup OmniWeb 5.1, allows remote attackers to read arbitrary files via the XMLHttpRequest Javascript component, as demonstrated using automatically mounted disk images and file:// URLs.

The International Domain Name (IDN) support in Safari 1.2.5 allows remote attackers to spoof domain names using punycode encoded domain names that are decoded in URLs and SSL certificates in a way that uses homograph characters from other character sets, which facilitates phishing attacks.

Apple Safari 1.2.4 does not obey the Content-type field in the HTTP header and renders text as HTML, which allows remote attackers to inject arbitrary web script or HTML and perform cross-site scripting (XSS) attacks.