This document proposes monitoring recommender systems over time to detect sybil attacks. It suggests distrusting newcomers to force sybils to draw out their attacks. It recommends monitoring at the system, user, and item levels by learning normal temporal behavior and flagging anomalies. The key contributions are forcing sybils to reveal attacks through prolonged activity and monitoring a wide range of attacks over time.
The document discusses how Markov chains are used as the methodology behind PageRank to rank web pages on the internet. It provides an overview of key concepts, including defining Markov chains and stochastic processes. It explains the idea behind PageRank, treating each web page as a journal and measuring importance based on the number of citations/links to other pages. The PageRank algorithm models web surfing as a Markov chain and the steady-state probabilities of the chain indicate the importance of each page.
This document proposes monitoring recommender systems over time to detect sybil attacks. It suggests distrusting newcomers to force sybils to draw out their attacks. It recommends monitoring at the system, user, and item levels by learning normal temporal behavior and flagging anomalies. The key contributions are forcing sybils to reveal attacks through prolonged activity and monitoring a wide range of attacks over time.
The document discusses how Markov chains are used as the methodology behind PageRank to rank web pages on the internet. It provides an overview of key concepts, including defining Markov chains and stochastic processes. It explains the idea behind PageRank, treating each web page as a journal and measuring importance based on the number of citations/links to other pages. The PageRank algorithm models web surfing as a Markov chain and the steady-state probabilities of the chain indicate the importance of each page.