This SoK paper systematizes the security landscape of mobile on-device AI (MoAI), where locally deployed models interact with conventional mobile software components. It organizes the field around security pillars, attacks, and defenses, while examining risks created by storing models on end-user devices. The authors argue that MoAI improves privacy, latency, and offline availability, but expands the local attack surface. The paper also identifies unresolved gaps and future research directions. Companion resources are provided in the linked Awesome-MoAI-Security repository.
No heat snapshots are available in the last 24 hours.