Skip to main content

Covertly intercepting video signals for espionage purposes may seem like a relic of the past, but a new discovery by researchers in Uruguay has introduced a frighteningly modern twist. A team from the University of the Republic in Montevideo has found that it’s possible to intercept the wireless electromagnetic radiation emitted by HDMI cables and interpret the video data using artificial intelligence (AI). This groundbreaking research was shared earlier this year on Cornell’s ArXiv platform, and it highlights how AI can be used to decode video signals without direct physical access to the target system.

The research explains that, despite HDMI being a wired connection, the electromagnetic signals emanating from the cable are strong enough to be detected from a distance. Even though HDMI signals are often encrypted using Digital Content Protection (HDCP), the radiation emitted by the cables contains enough information to be intercepted and decoded. With the help of AI models trained to process these electromagnetic fluctuations, the researchers demonstrated that it’s possible to interpret the data with up to 70 percent accuracy. This is a substantial improvement over earlier methods, which had an accuracy rate of only about 10 percent.

While 70 percent accuracy might not sound perfect, it’s more than enough to glean sensitive information like passwords or other personal data. In some cases, it’s even possible to intercept and decode the signals wirelessly from outside a building, making it a potential tool for surveillance without needing to be physically near the target.

This type of electromagnetic signal interception, known as TEMPEST (Transient ElectroMagnetic Pulse Emanation STandard), isn’t entirely new. It has its roots in espionage dating back to World War II. However, HDMI cables—particularly with their use of encryption like HDCP—weren’t thought to be vulnerable to this type of attack. The new AI-assisted technique, which the researchers have named “Deep-TEMPEST,” opens up new possibilities for wireless spying and surveillance.

The researchers suggest that this method, or similar ones, are already being used by state-level spies and agents engaged in industrial espionage. While the technique requires being in close proximity to the target system, it is unlikely to be a threat for average users. However, government agencies and companies dealing with sensitive data should be concerned. They may want to consider employing electromagnetic shielding measures, especially for employees working remotely, to mitigate the risk of being targeted by such high-tech spying methods.