"The amount and variety of content being published online is growing at an exceptional rate. Online publishing enables content to reach a much larger audience than paper publishing but offers no guarantee of long-term access to the content. This work investigates techniques for building a large, reliable peer-to-peer system for the preservation of online published material. The system consists of a large number of low-cost, persistent web caches (peers) that cooperate to detect and repair damage by voting in ""opinion polls"" on the content of their cached documents. The peers are autonomous and mutually suspicious. Project activities include 1) investigating defenses against adversaries whose goal is to attack the preservation process; 2) performing a foundational study of the interconnections between identity, trust, and reputation models in peer-to-peer systems; 3) investigating the use of estimates of peer diversity to increase the fault and attack tolerance of peer-to-peer systems; and 4) developing, analyzing, implementing, and testing new protocols that address the high frequency of updates of online government documents, the large volumes of scientific data, and the privacy concerns of sensitive medical data.This work is being evaluated using a real testbed of over 200 libraries around the world with the support of publishers representing over 2000 titles. The broader impact of the work is that all electronic material preserved through the system including academic journals, government documents and web articles, and scientific and medical data will remain accessible to generations of citizens for both research and education purposes."