Harnessing AI to Exploit HDMI Emissions

Researchers in the US have made a groundbreaking discovery by harnessing the power of deep learning to recreate digital video displays from the emissions originating from HDMI cables. The team at Cornell University delved into the realm of electromagnetic waves that are inadvertently emitted from cables and connectors, with a specific focus on HDMI technology.

When comparing the digital case of HDMI to the analog case of VGA, the researchers encountered significant challenges. The 10-bit encoding in HDMI leads to a substantially larger bandwidth and a non-linear mapping between the observed signal and the pixel intensity, making it a more complex scenario. Traditional eavesdropping systems designed for analog signals struggle to produce clear and coherent images when applied to digital video due to these complexities.

To tackle this issue, the team reframed it as an inverse problem and developed a deep learning module capable of mapping the observed electromagnetic signal back to the displayed image. Their innovative system is built upon the widely available Software Defined Radio known as TempestSDR and is fully open-source, seamlessly integrated into the popular GNU Radio framework.

Despite the advancements, this approach still necessitates a meticulous mathematical analysis of the signal. This analysis is crucial not only for determining the optimal frequency for tuning but also for generating the necessary training samples. By doing so, the researchers save valuable time and eliminate the need to acquire these samples manually, especially when exploring multiple configurations.

The team's ongoing efforts are focused on enhancing the average Character Error Rate in text. Their current system has demonstrated a remarkable improvement, surpassing previous implementations by over 60 percentage points. Additionally, they have generously shared the dataset created for training purposes, which includes a mix of simulated data and over 1000 real captures. Furthermore, the researchers are actively exploring countermeasures to mitigate the potential risks associated with eavesdropping systems that operate on similar principles.