Hot Pixels: Frequency, Power, and Temperature Attacks on GPUs and ARM SoCs

Hot Pixels is a significant academic research project (published in 2023) that demonstrated how software can exfiltrate sensitive data by measuring the frequency and power fluctuations of modern processors. Despite its technical depth, its defensibility as a project is low because it is a research PoC intended for disclosure rather than a commercial tool. With 0 stars and limited fork activity, it functions as a static reference for the paper's findings. The threat of platform domination is high because the 'fix' for such attacks must occur at the silicon or driver level by vendors like Apple, NVIDIA, and ARM. Once these vendors implement mitigations (e.g., coarsening timers or limiting access to frequency data), the utility of this specific implementation vanishes. It is conceptually similar to other GPU side-channels like 'GPU.zip' but specifically focuses on the thermal/power envelope. For a technical investor, the value here is in the domain expertise of the authors, not the repository itself.