Abstract
Graphics processing units (GPUs) on modern computers are susceptible to electromagnetic (EM) side-channel attacks that can leak sensitive information without physical access to the target device. Website fingerprinting through these EM emanations poses a significant privacy threat, capable of revealing user activities from a distance. This paper introduces EMMasker, a novel software-based solution designed to mitigate such attacks by obfuscating the EM signals associated with web activity. EMMasker operates by generating rendering noise within the GPU using WebGL shaders, thereby disrupting the patterns of EM signals and confounding any attempts at identifying user online activities. Our approach strikes a balance between the effectiveness of obfuscation and system efficiency, ensuring minimal impact on GPU performance and user browsing experience. Our evaluation shows that EMMasker can significantly reduce the accuracy of state-of-the-art EM website fingerprinting attacks from an average accuracy of 81.03% to 22.56%, without imposing a high resource overhead. Our results highlight the potential of EMMasker as a practical countermeasure against EM side-channel website fingerprinting attacks, enhancing privacy and security for web users.
Introduction
Modern GPUs are highly active in the electromagnetic spectrum. As a GPU renders a webpage, its power management systems—specifically Dynamic Voltage and Frequency Scaling (DVFS)—adjust clock speeds and voltages to match the workload. These fluctuations create unique EM patterns (emanations) that can be captured by a nearby attacker using a Software Defined Radio (SDR) and a directional antenna, even through walls. By training machine learning models like ResNet on these EM "heartbeats," attackers can perform website fingerprinting: identifying exactly which site a user is visiting without touching network traffic. While physical shielding is a possible solution, it is impractical for most consumers. EMMasker addresses this by providing a flexible, software-level defense.
Methodology: The EMMasker Framework
EMMasker is implemented as a browser extension that acts as a signal jammer for the GPU. The system follows a five-step operational workflow to protect user privacy:
- Triggering: The user specifies sensitive URLs to be protected. When the user navigates to these sites, the extension activates.
- Noise Generation: It initiates a hidden WebGL canvas. WebGL allows JavaScript to execute code directly on the GPU. EMMasker runs specialized fragment shaders designed to perform intensive, randomized rendering tasks.
- Dynamic Obfuscation: The system communicates with a local server to monitor the GPU's real-time clock frequency. If the GPU clock speeds up, EMMasker increases the intensity of the rendering noise to ensure the signal remains buried.
- Efficiency Management: To prevent overheating or lag, EMMasker triggers noise at random intervals (every 1–15 seconds) for short bursts (1–5 seconds), balancing privacy with usability.
Experimental Evaluation
The researchers evaluated EMMasker against a state-of-the-art attack model using a dataset of five high-traffic sites: Bing, Yahoo, Facebook, Google, and Amazon. Over 350 EM signal traces were collected for each website to train and test a ResNet-based classifier. The evaluation compared the classifier's performance on "clean" signals versus signals captured while EMMasker was active.
Key Results and Discussion
The defense proved highly effective. Without EMMasker, the ResNet classifier identified visited websites with an average accuracy of 81.03%. With EMMasker active, this accuracy plummeted to 22.56%, which is near the level of random guessing for a five-site dataset.
As shown in the signal traces, EMMasker works by forcing the GPU into higher performance states (P-states) for brief periods. These high-amplitude bursts completely overwrite the subtle EM patterns generated by normal web rendering. Despite this intensive rendering, the impact on user experience was marginal, with memory utilization increasing by only 7% to 13%.
Conclusion
EMMasker demonstrates that software-defined shielding is a viable path for protecting against hardware-level side-channel attacks. By using standard web technologies like WebGL to inject noise, it turns the attacker's target (the GPU) into a defense mechanism. While future work is needed to test a wider array of hardware and "open-world" website sets, EMMasker represents a significant step toward practical, consumer-ready privacy protections against long-range EM eavesdropping.
Summary updated