No abstract

Current Wi-Fi network infrastructure inherently lacks reliable positional knowledge of the origin of individual network packets. As a consequence, attackers are potentially able to impersonate legitimate Wi-Fi network nodes, including both clients and access points, by generating network traffic with all the characteristics of legitimate traffic. To exclusively discriminate illegitimate nodes, we propose a method of using multiple and multiple-angle measurements of a Wi-Fi network node's signal strength. In this paper we present a real time unsupervised analysis method for identifying duplicate Wi-Fi network nodes which are physically separate. We demonstrate that, given a dense sensor network, data from a small set of identifiable critical sensors are much more valuable in node discrimination than those from other sensors, enabling large improvement in responding speed and discrimination accuracy.

By design, Wi-Fi networks inherently lack reliable positional knowledge of the origin of individual network packets, and instead use the media access control (MAC) address of a received packet to assign packet identity. Given that there are a variety of tools that allow one to construct arbitrary network packets and transmit over a network, a potential security risk arises in that it is difficult to discriminate legitimate packets from packets generated by an intruder solely on the basis of packet content. We describe a data clustering approach to discriminating network nodes based on signal strength data measured from multiple angles through a distributed sensor network, and demonstrate that relatively little network data are sufficient for high discrimination accuracy.

No abstract