Researcher: Radu Marculescu
While there has been considerable research on secure routing, information forensics and trusted computing technology against various attacks, there is no research targeted at risk analysis. Risk analysis is an analytical technique meant to identify and assess the factors that can jeopardize network security. More precisely, we propose to use this technique as a predictor for ensuring the desired level of security in a given network. Consequently, the risk analysis becomes a useful design metric which can help design an anonymous routing protocol by reducing the vulnerability to various detrimental factors. In this research, we will address the risk analysis problem under eavesdropping risk, which can be defined as, "Given an arbitrary user distribution and w adversarial nodes present in an ad-hoc wireless network, find the distribution of adversaries such that the probability of packets being eavesdropped is maximized."
Obviously the eavesdropping risk problem depends on the number of adversaries, the underlying wireless communication technology (e.g. transmission power control), locations of users, traffic patterns among users and user mobility. For the sake of tractability,we start by analyzing the first-order eavesdropping risk problem and then extend the study to higher orders. We also plan to develop a simulator in order to assess the risk whenever the mathematical analysis turns out to be intractable. The main steps are summarized below:
Analyze the Impact of Transmission Power Control on the Eavesdropping Risk
We plan to use the Gammal's model to capture the geographical structure and interference properties of the ad-hoc wireless networks. The Gammal's model simplifies the analysis process, making it possible to derive an analytical relationship between the transmission power control and the eavesdropping risk. The results of this study will thus verify the usefulness of transmission power control on network security.
Design the User Geographical Distribution
We plan to take two parallel approaches and consider both random and deterministic user distributions. For simplicity, we will start with the uniform user distributions; that is, each user node is assigned a location independently and uniformly over the entire deployment area. We will then generalize the user geographical distribution and study how the user distribution affects network security.
Effects of User Mobility on Eavesdropping Risk and Latency
This issue is critical because a routing protocol with strong dependence on mobility may cause a huge overhead and overlong latency. These adverse effects are not acceptable for real-time traffic transmission. On the other hand, the knowledge of tradeoffs and synergies between mobility, security and latency can help the design of lighter routing protocols, while satisfying all the application specific requirements. Quantitative analysis of possible tradeoffs and synergies and the development of a simulation environment for evaluation are the main objectives of this research.