| Summary: | The number of devices connected to the Internet has been growing exponentially over the last few years. Today, the amount of information available to users has reached a point that makes it impossible to consume it all, showing that we need better ways to filter what kind of information is sent our way. At the same time, while users are online and access all this information, their actions are also being collected, scrutinized and commercialized with little regard for privacy. This thesis addresses those issues in the context of a decentralized Publish/Subscribe solution for edge systems. Working at the edge of the Internet aims to prevent centralized control from a single entity and lessen the chance of abuse. Our goal was to devise a solution that achieves efficient message delivery, with good load-balancing properties, without revealing its participants subscription interests to preserve user privacy. Our solution uses cryptography and probabilistic data sets as a way to obfuscate event topics and user subscriptions. We modeled a cooperative solution, where publisher and subscriber nodes work in concert to route events among themselves, by leveraging a onehop structured overlay. By using an experimental evaluation, we attest the scalability and general performance of the proposed algorithms, including latency, false negative and false positive rates, and other useful metrics.
|
|---|