Security researchers have found that thermal cameras can be combined with computer algorithms to automate the process of stealing payment card data processed by automatic teller machines.
At the Usenix Security Symposium in San Francisco last week, the researchers said the technique has advantages over more common ATM skimming methods that use traditional cameras to capture the PINs people enter during transactions. That’s because customers often obscure a camera’s view with their bodies, either inadvertently or on purpose. What’s more, it can take a considerable amount of time for crooks to view the captured footage and log the code entered during each session.
Thermal imaging can vastly improve the process by recovering the code for some time after each PIN is entered. Their output can also be processed by an algorithm that automates the process of translating it into the secret code.
The findings expand on 2005 research from Michal Zalewski, who is now a member of Google’s security team. The Usenix presenters tested the technique laid out by Zalewski on 21 subjects who used 27 randomly selected PINs and found the rate of success varied depending on variables including the types of keypads and the subjects’ body temperature.
“In summary, while we document that post-hoc thermal imaging attacks are feasible and automatable, we also find that the window of vulnerability is far more modest than some feared and that there are simple counter-measures (i.e., deploying keypads with high thermal conductivity) that can shrink this vulnerability further still,” the researchers wrote.
A PDF of their paper, which is titled Heat of the Moment: Characterizing the Efficacy of Thermal Camera-Based Attacks, is here.