Enforcing Benign Trajectories: A Behavioral Firewall for Structured-Workflow AI Agents

arXiv:2604.26274v1 Announce Type: cross Abstract: Structured-workflow agents driven by large language models execute tool calls against sensitive external environments. We propose odename, a telemetry-driven behavioral anomaly detection firewall. Drawing on sequence-based intrusion detection, odename compiles verified benign tool-call telemetry into a parameterized deterministic finite automaton (pDFA). The model defines permitted tool sequences, sequential contexts, and parameter bounds. At runtime, a lightweight gateway enforces these boundaries via an O(1) state-transition structural lookup, shifting computationally expensive analysis entirely offline. Evaluated on the Agent Security Bench (ASB), odename achieves a 5.6% macro-averaged attack success rate (ASR) across five scenarios. Within three structured workflows, ASR drops to 2.2%, outperforming Aegis, a state-of-the-art stateless scanner, at 12.8%. odename achieves 0% ASR on multi-step and context-sequential attacks in structured settings. Furthermore, against 1,000 algorithmically spliced exfiltration payloads, only 1.4% matched valid structural paths, all of which failed end-to-end string parameter guards (0 successes out of 14 surviving paths, 95% CI [0%, 23.2%]). odename introduces just 2.2~ms of per-call latency (a 3.7imes speedup over extsc{Aegis}) while maintaining a 2.0% benign task failure rate (BTFR) on benign workloads. Modeling the behavioral trajectory effectively collapses the available attack surface, but unmaintained continuous parameter bounds remain vulnerable to synonym-substitution attacks (18% evasion rate). Thus, exact-match whitelisting of sensitive parameters ultimately bears the final defensive load against execution.

Source: arXiv cs.AI | 2026-04-30