Deep Learning Enables Eavesdropping on Digital Video Displays

Eavesdropping on digital video displays through electromagnetic emanations: Researchers have developed a deep learning-based system called DEEP-TEMPEST that can effectively eavesdrop on digital video displays, such as HDMI, by analyzing the unintentional electromagnetic waves emanating from cables and connectors:

• The digital case, particularly HDMI, poses a greater challenge compared to analog (VGA) due to a 10-bit encoding that results in a larger bandwidth and a non-linear mapping between the observed signal and pixel intensity.
• Existing eavesdropping systems designed for analog video obtain unclear and difficult-to-read images when applied to digital video, necessitating a new approach.

Deep learning as a solution: The researchers recast the problem as an inverse problem and trained a deep learning module to map the observed electromagnetic signal back to the displayed image:

• This approach requires a detailed mathematical analysis of the signal to determine the frequency at which to tune and to produce training samples without needing a real TEMPEST setup.
• By generating training samples computationally, the researchers saved time and avoided the need to obtain these samples physically, especially when considering several configurations.

Improved performance and open-source implementation: The proposed DEEP-TEMPEST system significantly improves the average Character Error Rate (CER) in text compared to previous implementations and is based on widely available hardware and software:

• DEEP-TEMPEST improves the CER by over 60 percentage points compared to previously available implementations, making it more effective at eavesdropping on digital video displays.
• The system is built using widely available Software Defined Radio (SDR) and is fully open-source, seamlessly integrated into the popular GNU Radio framework.
• The researchers also shared the dataset they generated for training, which includes both simulated and over 1,000 real captures, enabling further research and development in this area.

Countermeasures and potential risks: The researchers discuss countermeasures to minimize the potential risk of being eavesdropped by systems designed based on similar principles:

• The development of DEEP-TEMPEST highlights the potential vulnerability of digital video displays to eavesdropping through electromagnetic emanations.
• While the researchers propose countermeasures, the open-source nature of the system and the shared dataset may enable malicious actors to develop similar eavesdropping tools, raising concerns about the security of sensitive information displayed on digital video screens.

Broader implications for digital security: The development of DEEP-TEMPEST underscores the ongoing need for robust security measures in the digital age, as even seemingly secure technologies like HDMI can be vulnerable to eavesdropping through unintended electromagnetic emanations. As deep learning techniques continue to advance, it is crucial for researchers and security experts to proactively identify and address potential vulnerabilities in digital systems to protect sensitive information from unauthorized access.