AI and the World's Hard Problems · Quantum physics & quantum computing · arXiv · 2026-10-20
Q-MALDetect: Quantum Image-Based Malware Detection and Adversarial Robustness Analysis
Q-MALDetect is a hybrid convolutional and variational-quantum detector for byte-image representations of ELF binaries, tested against feature-space and executable-aware evasion attacks. It reached 97.2% training accuracy, but its robustness fell as attack strength rose, with the strongest feature-space attack achieving a 78.7% success rate. Dynamic adversarial training was the most effective defense, lowering feature-space attack success to as low as 5.0%.