A concerning vulnerability has been revealed in modern computer systems, as per a new study. Federico Larroca at the University of the Republic in Montevideo, Uruguay, and his colleagues have developed an AI model that can reconstruct digital signals intercepted from HDMI from a few metres away.

By capturing electromagnetic radiation emitted from the cable, the AI model can accurately decipher a significant portion of the displayed content. Attacks that utilize this technique are known as TEMPEST. The research team used text capture software on the recovered image and compared the resulting text to the original screen image. While the current error rate is around 30%, it is significantly lower than previous methods, indicating that the error rate will likely go down even more in the future.

The technique relies on the principle that digital signals, while more complex than analog ones, still leak electromagnetic radiation during transmission. Hackers could intercept sensitive information such as passwords, financial data, or confidential documents. It's easy too - one could be standing outside a building with an antenna that captures images in real time, or they could plant a device that captures and transmits (or stores) images.

While the researchers explicitly state that the average user is unlikely to be targeted, organizations handling sensitive information should be aware of this threat and implement appropriate countermeasures. According to Larroca, it is probable that such attacks are already occurring in highly sensitive industrial/government settings. Hacks and breaches have been on the rise lately, for instance, WazirX losing around $230 million to a crypto scam on July 19.