AI & ML Practical Magic

Hackers can now 'see' your screen from a distance just by looking at how light bounces off the wall next to it.

April 16, 2026

Original Paper

Physically-Guided Optical Inversion Enable Non-Contact Side-Channel Attack on Isolated Screens

Zhiwen Zheng, Yuheng Qiao, Xiaoshuai Zhang, Zhao Huang, Tao Zhang, Huiyu Zhou, Shaowei Jiang, Jin Liu, Wenwen Tang, Xingru Huang

arXiv · 2604.13419

The Takeaway

This paper introduces IR4Net, a neural network that can reconstruct a hidden screen's content by analyzing the 'speckle patterns' of diffuse light reflections. You don't need a direct line of sight; just the light hitting a nearby surface is enough. This turns every wall or object near a monitor into a potential security leak. It represents a massive leap in side-channel attacks, moving from theoretical curiosities to a high-fidelity reconstruction capability. For security practitioners, this changes the threat model for 'air-gapped' systems. You now have to worry about optical leakage from surfaces you previously thought were safe.

From the abstract

Noncontact exfiltration of electronic screen content poses a security challenge, with side-channel incursions as the principal vector. We introduce an optical projection side-channel paradigm that confronts two core instabilities: (i) the near-singular Jacobian spectrum of projection mapping breaches Hadamard stability, rendering inversion hypersensitive to perturbations; (ii) irreversible compression in light transport obliterates global semantic cues, magnifying reconstruction ambiguity. Explo