Robust detection of malicious behavior in distributed wireless networks.

点击翻译按钮获取中文摘要

This project aims at developing a better understanding of security architectures for wireless networks, with particular focus on wireless sensor networks (e.g. in industrial environments for detecting malfunctions in production processes) and cognitive radio networks, taking into account the unique features and limitation of these systems. The core premise of the proposal is that efficient on-line malware detection algorithms must account for possibly unreliable or inaccurate information due to transmission errors or faulty devices, and optimize the utilization of scarce available wireless resources, including an appropriate prioritization of protected network elements. The first step towards this goal is the development of a general statistical framework, which shall encompass a number of different malfunctions, attacks, and anomalous behaviors in wireless networks. This model will serve as a starting point and common background for the application of a variety of statistical inference and machine learning techniques. The analysis will be then complemented by considering the anomaly detection problem jointly with the resource allocation problem, in an effort to conjugate security with system-level network optimization. Thus, statistical detection theory will be combined with optimization theory, game theory and information theory. Finally, the high variability of data provided by individual sensor nodes naturally calls for distributed solutions, especially in large-scale networks. As some of the devices actively involved in the collaborative decision process might be malfunctioning or under the control of malicious users, the project will investigate incentive mechanisms and selective cooperation schemes based, for instance, on estimated reputation values assigned to each node.