A parameterized DFA firewall enforces safe tool sequences for structured AI agents, reducing attack success rates to 2.2% in tested workflows with low added latency.
Communications of the ACM19(5), 236–243 (1976) 28
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
cs.CR 1years
2026 1verdicts
UNVERDICTED 1representative citing papers
citing papers explorer
-
Enforcing Benign Trajectories: A Behavioral Firewall for Structured-Workflow AI Agents
A parameterized DFA firewall enforces safe tool sequences for structured AI agents, reducing attack success rates to 2.2% in tested workflows with low added latency.