Research Mentor: Lei Wang
The objective of this project is to develop a probabilistic analysis framework that allows designers to study the security vulnerability in embedded systems. A variety of reports have shown that embedded systems are vulnerable to many forms of attacks, including software viruses that allow the implantation of backdoors, side channels that can acquire security-associated properties by non-invasively analyzing system behaviors, and hardware tampering that exploits system implementation details such as inter-component communications. Most existing research focuses on secure data processing and communication in embedded systems. However, embedded systems feature heterogeneous interfaces to numerous devices, as well as interactions with different users. These features require complicated control modules at all levels of system hierarchy to ensure proper system responses and functionalities, which have not received much attention so far. Attacks on control modules are relatively easy to carry out with cheap equipment. For example, random fault injection can be launched by under-powering a device, tampering clock signals, or exposing a device to intensive radiation. Targeted attacks can also be performed through side channels with the knowledge of system design and assistance of data analysis. The project will develop a statistical evaluation method, which provides a reliable measure on the security risks of embedded system under various attacks. Furthermore, the proposed framework also delivers easy-to-implement solutions toward a more robust and secured control system architecture.
The REU students are expected to develop a set of attack models for target embedded systems. This task is pivotal for building an evaluation framework that identifies the vulnerabilities in embedded systems. While the general mechanisms of random or targeted fault injection based attacks are known, the dynamic responses of embedded systems and their control state transitions in the presence of these attacks are not well-understood. This task will define the basic scope of attack model formalization. The effectiveness of the attacks will be quantified by analytical metrics, such as the probability of maliciously altered inner state transitions, and the security loss under various attack scenarios will be determined with respect to the associated data assets. Likelihood/consequence-based measures will be derived as the output of these models.