Sai Manoj Pudukotai Dinakarrao, Assistant Professor, Electrical and Computer Engineering, is developing a technique to detect and defend against hardware Trojans (HT) in integrated circuits (ICs) by adapting techniques inspired by the human immune system and vaccination.
The mechanism he and his collaborators propose includes an irregular network topology for connecting different macro or IP blocks similar to microbial colonies, and cortical interconnections, as depending on the design principle, these are found to display a striking resilience against random failures and targeted attacks.
The researchers will use a particular class of complex networks called small-world networks that are characterized by many short-distance links between neighboring nodes as well as a few relatively long-distance direct shortcuts.
In addition to system-level design, the researchers will also introduce a human immunology-inspired HT-triggering mechanism and defense mechanism. The system is inspired by the human immune system, where the body is forced to generate antibodies even when an antigen has not caused infection. This is performed by vaccination, wherein a set of benign antigens are introduced in the body that will force the generation of antibodies. When an actual antigen later infects the body, the generated antibodies will fight against them.
Inspired by this, the researchers propose a hardware Trojan detection mechanism. Once the researchers obtain an IC from a foundry, it is passed under abnormal conditions, e.g., overclocking. A rogue foundry will insert a Trojan on the assumption that the IC operates in a normal condition. Under these conditions, a Trojan can hide its effects within the IC operation. The Trojan doesn't prepare itself for conditions which are not deemed normal; and hence, the reaction of a Trojan-infected circuit will be completely different from one where a Trojan is not present. Thus, subjecting the monitored IC to unnatural conditions will help to detect the Trojan.
To ensure the impact of HTs is present, the researchers will deploy a machine learning-based anomaly detector capable of collaborative learning that is trained with the operational behavior of an IC under the normal conditions as well as HT presence from different vendors without violating the security and privacy of the systems. Thus, the operational behavior of an IC under the stimulated environment will be analyzed for efficient HT detection.
The proposed mechanism can be deployed in parallel on the existing and emerging Air Force C4I processing systems as well as other U.S. Department of Defense warfare devices and other commercial systems.
Pudukotai Dinakarrao received $19,950 from the Defense Advanced Research Projects Agency (DARPA) for this project. Funding began in August 2020 and will end in February 2021.
###