Research On Data Availability And Security In Peer-to-Peer Storage System

Peer-to-Peer (P2P) storage systems have been one of the hottest research issues for its attractive advantages, such as self-organizing, scalability and fault tolerance. However, data availability and security are two serious problems because of the high dynamic and untrusted peers. Though a number of research efforts have focused on these issues, many problems remain unsolved, which set great barrier to the real deployment.As for the data availability, this thesis first measures the dynamics of P2P systems, and then proposes schemes to ensure desired availability by using minimal bandwidth in 2 steps: (a) By using a fine-grained analysis model and placement strategy, we make the minimal data redundancy to mask transient leaves; (b) By accurately estimating the number of permanent departures, we use the minimal system bandwidth to repair the lost data redundancy. As for the security, this thesis addresses and analyzes the new confidentiality problems in P2P storage environment, and presents a new scheme to solve the problems.The main contributions of this dissertation includes,(1) Measurements of real deployed P2P environments. This thesis makes a thorough measurement and comparison of the dynamic natures of different types of P2P environments, which support the research of this thesis as well as others. This thesis is the first work to reveal the inherent limitation of crawler-based measurement, to model the online time distribution of a single peer, to find out the high dynamic characteristic of the new coming peers.(2) A fine-grained availability definition and analysis model, and placement strategy based on peer dynamics. This thesis is the first to show that the desired availability cannot be guaranteed by the conventional random placement strategy. By presenting a time-related availability definition and its analysis model, this thesis develops a dynamics based placement strategy. Experiments and analyses show that the new strategy guarantees the desired availability by using much less redundancy.(3) PROTECTOR, a permanent departure detector based on the balance of false positive and false negative. The time-threshold detector is widely used in permanent departure detection, but it cannot reduce the false positive and the false negative at the same time, and it has the difficulty in choosing the threshold. PROTECTOR balances the false positive and false negative by using a novel probability scheme, so it can accurately estimate the number of departed peers. Experiments show that PROTECTOR saves much bandwidth. Extra replicas are widely used to avoid the repair triggered by transient leaves, and this thesis is the first to point out that they cannot actually save the bandwidth.(4) SEC, a secure erasure code scheme for P2P storage environments. This thesis addresses the adversary model and the security analysis model for P2P storage environments. The analysis shows that the conventional block cipher scheme has potential vulnerabilities, and that the SEC scheme can solve the problems.(5) UPStore, an open P2P storage service platform. This thesis designs the UPStore system, which can provide high available and high secure data storage service by adopting all research results in this dissertation.