Abstract
The iTrust search and retrieval network is designed to impede attempts to censor or filter information accessed over the Internet, providing trustworthy access to information on the Web. In iTrust, a fully distributed membership algorithm and a detection/defensive adaptation algorithm act in concert to protect against malicious nodes in membership. By applying statistical inference for both detection and defensive algorithms, I was able to measure information that cannot be observed directly, such as the current size of the membership and the current proportion of malicious nodes in the network. Experimentation demonstrates that both algorithms are able to estimate these metrics quickly and accurately, to the point that the nodes can use them to manage the iTrust system, despite a high rate of membership churn, a large number of malicious nodes, and a mere partial view of network membership.
Original language | English |
---|---|
Pages (from-to) | 3492-3510 |
Number of pages | 19 |
Journal | Security and Communication Networks |
Volume | 8 |
Issue number | 18 |
DOIs | |
State | Published - 1 Dec 2015 |
Keywords
- Distributed system
- Dynamic adaptation
- Membership churn
- Search and retrieval
- Statistical inference
- Trustworthiness