Protecting documents from cross-site scripting attacks

Abstract

In various implementations, an embedded document receives untrusted content from a containing document, where the embedded document is in the containing document. In some cases, the untrusted content is received by the containing document from a server and is forwarded to the embedded document without rendering the untrusted content in the containing document. Instead, the untrusted content is rendered in the embedded document. A sandbox policy is enforced on the embedded document such that the rendered untrusted content is restricted from accessing data associated with the containing document. The untrusted content may comprise malicious code that when rendered executes an XXS attack that attempts to access the data associated with the containing document. However, because the untrusted content is rendered in the embedded document, the malicious code may be denied access to the data, thereby preventing the XSS attack from succeeding.