Internet traffic is continuously growing and contributes substantially to rising costs for network operators. Evaluations have shown that today multimedia content accounts for a major part of the transferred bytes in the Internet and that HTTP is the dominant protocol. A natural solution for reducing these network costs is caching of frequently requested content. Already in the beginning of the 90s HTTP caches have been proposed, which were deployed in the domains of the network operators. These traditional HTTP caches rely on URLs to identify resources and to avoid transferring the same data twice. Unfortunately today a specific content might be available under different URLs. Furthermore many HTTP connections are personalized and therefore caching is often disabled by content producers. So traditional HTTP caching became inefficient for the network operators. In this paper we propose a method to improve the efficiency of HTTP caching. Our approach is based mainly on hash keys as additional identifiers in the header of HTTP messages. By that identification of the transferred content is more precise than with URLs. Beside this we show how caching can be achieved even in the presence of personalization in HTTP messages and how content producers remain full control over their content although it is cached by ISPs.