Network congestion remains a major concern to WWW users. This congestion manifests itself in unacceptably long response times. One possible remedy to the latency problem is to use caching at the client, at the proxy server, or even within the Internet. However, WWW documents are increasingly dynamic (i.e., have short lifetimes), which limits the potential benefit of caching. The performance of a WWW caching system can be dramatically increased by combining it with prefetching (a.k.a., proactive caching). Although prefetching reduces the response time of an interactive Web user, it generates additional network traffic and consumes buffering resources when the prefetched documents end up not being requested by clients. Hence, the degree of prefetching should be adjusted, depending on the level of confidence in the predicted future demand, the available buffering resources at the proxy/client caches, and the network traffic intensity. Our research is focused on: (1) developing stochastic models for characterizing the behavior of WWW traffic, and (2) using such models in designing, analyzing, and optimizing the performance of proactive caching protocols. For the modeling part, we have used multifractal analysis as a means of characterizing the irregularities in web traffic at multiple time scales.