Study Reveals Hackers Can Steal Data Via HDMI Cables
Morrissey Technology – A recent study by a research team at the Universidad de la Republica Montevideo, Uruguay, revealed that hackers can use artificial intelligence (AI) and HDMI cables to steal data on their victims’ computers. Read the explanation. The research, entitled ‘Deep-TEMPEST: Using Deep Learning to Eavesdrop on HDMI from its Unintended Electromagnetic Emanations’, was published on the arXiv preprint server. The research was conducted by Santiago Fernandez Emilio Martinez, Gabriel Vareal, and Pablo Muse Federico Larroca.
HDMI cables are usually used to transfer the screen display from a laptop or PC to a projector to have a larger display. This cable apparently produces electromagnetic radiation that can be captured by hackers. In this study, researchers found that hackers also used AI technology in hacking practices via HDMI cables.
The study involved capturing electromagnetic radiation emitted from a computer’s HDMI cable. They then trained the AI system by providing screen samples associated with radiation signals traveling through the computer’s HDMI cable. As time went by, these systems gradually became better at interpreting text displayed on computer screens.
Test results show this system is able to reconstruct text from a computer screen randomly with 70 percent accuracy. They also note these systems may be good enough to steal passwords, sensitive data, or in some cases, encrypted communications. The researchers found that they could improve their results by using text recognition software on the text after it had been parsed.
Researchers say hackers may have conducted similar research. This means such screen hacks may have targeted a number of victims. According to them, all that is needed is hardware capable of capturing the electromagnetic radiation emitted by HDMI cables placed close to buildings, such as in the back seat of a car. The team further suggests that most people are not at risk of such an attack, due to the expertise required by this technique. Instead, they suggest that it is more likely that governments or corporate entities will be targeted.