by Joshua Timmons, Biology, 2017
Side-channel attacks are an emerging threat to embedded systems. Attackers measuring power fluctuations, or the time required to perform cryptographic algorithms, can reverse-engineer the implementation of the algorithm being executed rather than crack the algorithm itself.
Countermeasures against such side-channel attacks are in high-demand, in part from the growing use of smart-chips. Credit cards, identification cards (for door access), and even the Boston’s Charlie Card rely upon integrated chips for security. Whereas the magnetic strips of days passed could be cloned and reused for nefarious purposes, today’s integrated chips are considered a safe alternative that cannot be copied. However, to an expert hacker using a side-channel attack, even these embedded chips may be vulnerable to exploitation.
In light of this emerging threat, and the need for solutions, Associate Professor Adam Ding and colleagues at Northeastern, Yunsi Fei (ECE) and Thomas Wahl (CCIS), are working on a multidisciplinary project to automate the identification of device weaknesses to side-channel attack. With a recent grant, “Automating Countermeasures and Security Evaluation Against Software Side-channel Attacks,” they plan to create a tool that automatically corrects for side-channel weaknesses.
Ding’s work, in particular, has focused on developing a quantitative model that unifies the side-channel exploitation of two major forms of attack: power-usage and cache timing. A statistician, Ding quantifies the information contained in data to quantify the information leakage through side-channels. “Given any kind of secure parameter, from our model, we can calculate how much signal is required [to hack the device],” said Ding, “From a statistical view, what we can give is a quantification.”
This quantification is a prediction of how many operations would be needed to achieve a side-channel exploit and is directly relates to the vulnerability of the device. A non-leaky device will require a relatively high number of operations, have a low signal to noise ratio, and will be secure form a user standpoint. “If you specify a security requirement, how do we achieve and ensure that the side-channel leakage signal is lower than that?” said Ding.
Ding’s mathematical model is integrated with an engineering approach to automate the countermeasures on the software level of the device, with a computer science approach for formal verification. Cumulatively, Ding, Fei, and Wahl have set out to fully automate protection against side-channel leakage.
This tool would be particularly useful given the present technical difficultly of patching up security loopholes. “To do this manually, it is very sophisticated work,” said Ding. The goal of Ding and his interdisciplinary group is to enable a greater number of manufacturers to create side-channel protected devices, rather than the specialized few that are able to do so today. “The automation tools will allow much wider applications of side-channel protection, serving the societal need for secure computations and communications,” said Ding.
By predicting vulnerabilities using their model, and then automatically fixing them on a software level, Ding and colleagues hope to empower a greater number of manufacturers to protect against side-channel attacks.