Abstract
Centralized architecture, due to its simplicity, fast and reliable user management mechanism (authorization, authentication and lookup) and O(1) searching capability, is still a preferable choice for many P2P-based services. However, it suffers from a “single point of failure” vulnerability, so networks based on this topology are highly vulnerable to DoS attacks or other blocking attempts. This paper describes a new mechanism that can be used for centralized P2P networks to prevent a P2P service unavailability after central server failure. High security level is obtained by using notary servers which track server public key changes and collect social feedback from users. This allows not only to detect popular attacks (like man-in-the middle) but also to assess whether the Central Server (CS) behaves properly. In the case of central server failure or when server becomes compromised, decentralized Condorcet voting is preformed and new CS is selected. Additionally, by incorporating a reputation mechanism which uses two kinds of scores respectively for providing good service and fair evaluation of other peers, the best candidates for a new Central Server can be chosen. Valuable data which is used to rebuild user database in new CS is stored in the encrypted form in peers and updated during the user-peer authorization process. The decryption key is divided between peers using the threshold secret sharing method.
Highlights
In recent years, peer-to-peer systems have become more and more popular in many domains like file sharing, voice and video communication (e.g. Skype), streaming content distribution [1], as well as distributed storage and computing centers
S paper we present a novel self-healing mechanism that can be used for building systems that will incorporate both: the fast and reliable searching and user managing mechanism appropriate for the centralized P2P systems and high failure tolerance level as well as blocking resistance known from the decentralized infrastructures
In our case, even if infrastructure is taken over - by selecting a hostile peer for the Central Server, any deceitful behaviour of this new server will be detected by users and Notary Servers will be informed, which will result in repeating the voting process and selecting another Central Server
Summary
Peer-to-peer systems have become more and more popular in many domains like file sharing, voice and video communication (e.g. Skype), streaming content distribution [1], as well as distributed storage and computing centers. DHT provides a search mechanism similar to hash tables, which is used to associate certain resources (for example file ownership) to a certain peer This topology has many advantages like good scalability, satisfying fault tolerance and reliable search mechanism (typically O(log N ) messages to resolve a search query in the N nodes network) which makes it a good candidate for file sharing services. C ally, the presented method is easy to implement in the existing P2P systems with the central server, which will result in significant increase of security, while retaining the previous performance This objective has been achieved by incorporating three additional mechanisms, responsible respectively for central server misbehaviour detection, decentralized voting used for choosing a new server, and central database rebuilding (using encrypted backups stored in peers and secret sharing technique). Negative opinions about the CS operation are reported manually by clients and the whole mechanism works to the anti-phishing software [12]
Sign-in/Register to view highlights & summary Sign-in/Register to access full text options 
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have
Schedule a call
