Self-State Attacks on Self-Hosted AI Agents: How Far Can OS Defenses Go?
For developers and operators of self-hosted AI agents, this paper identifies a new class of attacks and demonstrates that current OS defenses are insufficient, motivating new research directions.
This paper introduces self-state attacks on self-hosted AI agents, where an agent's own memory and configuration files are corrupted via legitimate OS system calls. The authors characterize a four-axis attack space, evaluate OS-level defenses, and find that a layered defense stack is effective against most attacks, though a small residual attack surface remains structurally indistinguishable at the OS level.
Self-hosted AI agents read and write their own memory and configuration files to function. An agent may get compromised via corruption of its own state -- a compromise realized via legitimate OS system call invocation. We refer to this class of threats as self-state attacks. In this paper, we investigate the OS resilience to this class of attacks. Formally, we characterize a four-axis attack space (Target, Mechanism, Granularity, Temporal); investigate the structural limits of prevention, detection, and recovery; and introduce a workload-conditioned view of detectability. To instantiate the framework, we collect live activity traces from a representative self-hosted agent running across distinct workload profiles, and realize the attack space as a 23-cell matrix, 43 concrete operations on real self-state files, and injected into those traces. We then evaluate both canonical and workload-conditioned defense strategies. The empirical results show that a layered defense stack (access-control prevention on the instruction and configuration layers, workload-conditioned detection on the memory layer, and periodic backup for recovery) is effective on most attack cells while a small residual attack surface remains structurally indistinguishable at the OS level. These findings suggest that against the newly established class of self-state attacks, OS-level defense needs to be reconsidered, potentially opening new research directions in the field.