Security of many computer systems relies on the basic assumption that data theft through unauthorized physical tampering with a computer system is hard and easily detected. Unfortunately, the ongoing transition to a knowledge economy results in increasing motivation for attackers and results in increasingly sophisticated attacks. Potential EM side-channel attacks, which use radio and microwave signals that are unintentionally produced by computer hardware as it operates, are especially worrisome because they can be carried out without actual contact with the computer, and because they are virtually undetectable by the user of the system. Unfortunately, there is very limited understanding of how much information is possibly being leaked from modern computer systems, from which distance can this information be received, and how to design processors and systems in a way that systematically minimizes this data leakage.
This research proposes to carry out preliminary investigation and characterization of the EM side channel data leakage, and gain useful insights that will pave the way toward approaches that minimize this leakage. To this end, the PIs will create a basic experimental testbed to receive and analyze EM emanations from modern systems, and then use this testbed to identify EM emanations that leak data and estimate how the amount of EM data leakage changes with the distance from the system.
Security of most computers relies on the assumption that attacks must either exploit vulnerabilities in software to gain access, or break encryption used for communication to/from the system. However, a large class of attacks, collectively called "side-channel" attacks, extracts information by analyzing properties related to the physical behavior of the system, such as changes in power consumption, sounds created by the system, etc. Among these side-channel attacks, those that use electromagnetic (EM) signals (e.g. radio waves) emitted by the system are especially dangerous, because these signals can be received without physical contact with the computer. The goal of this NSF-funded project was to gain initial insights into which EM signals are generated by typical laptop and desktop computers, how far they propagate, and what types of information do they leak. We have successfully created a basic measurement system for collecting and analyzing the behavior of computer systems and the resulting EM emanations in the AM radio band, and have shown that specific activity patterns result in data-carrying emanations that propagate to distances of at least 10 feet from the system. These results will be used to initiate further research projects that will lead to understanding of how these signals are transmitted and how they can be prevented without resorting to heavy shielding or power-hungry noise generators.