The Web has evolved to support sophisticated web applications. These web applications are exposed to a number of attacks and vulnerabilities. The existing security model is unable to cope with these increasing attacks and there is a need for a new security model that not only provides the required security but also supports recent advances like AJAX and mashups. The attacks on client-side Web Applications can be attributed to four main reasons – 1) lack of a security context for Web Browsers to take decisions on the legitimacy of requests, 2) inadequate JavaScript security, 3) lack of a Network Access Control and 4) lack of security in Cross-Domain Web Applications. This work explores these four reasons and proposes a new security model that attempts to improve overall security for Web Applications. The proposed security model allows developers of Web Applications to define fine-grained security policies and Web Browsers enforce these rules; analogous to a configurable firewall for each Web Application. The Browser has disallows all unauthorized requests, thus preventing most common attacks like Cross-Site Script Injections, Cross-Frame Scripting and Cross-Site Tracing. In addition the security model defines a framework for secure Cross-Domain Communication, thus allowing secure mashups of Web Services. The security model is backward compatible, does not affect the current usability of the Web Applications and has cross-platform applicability. The proposed security model was proven to protect against most common attacks, by a proof-of-concept implementation that was tested against a comprehensive list of known attacks.