Security Via Collaborative Monitoring

Heuristic search for "least privilege / maximum functionality" program configurations can find the best way to achieve security for complex applications. This research project will draw on peer-to-peer networking and file sharing techniques to build huge computational networks that collect, collate and process data using distributed learning algorithms. The amount of computing power and the variety of data sources will make it possible to extend the learning algorithms beyond the current state-of-the-art, and they will be able to determine normal program behaviors and configuration parameters. The recommended security settings will be adjusted to common usage environments. The system will preserve anonymity and privacy, and the effects of malicious participants will be confined. The project will develop the tools and methods within a flexible, open framework.