Framework sets safe boundaries for autonomous cyber defense actions
SecODD: Adapting the Operational Design Domain from Driving Autonomy to Autonomous Cyber Defense
Cryptography and Security
Summary
Autonomous agents are starting to handle cybersecurity tasks automatically, but mistakes could cause major problems or disputes over responsibility. The authors propose SecODD, a framework that keeps these actions within safe limits, ensures humans can always intervene, and makes all steps traceable for audits. It borrows ideas from self-driving car regulations to clearly define when and how an autonomous system can act. This approach helps prevent errors and supports accountability in automated cyber defense.
What this means in practice
- •For security operations teams: Manage and govern automated cyber defense tools to ensure safe, auditable, and interruptible responses in live security incidents.
- •For managed security service providers: Implement oversight frameworks that define clear operational boundaries for outsourced autonomous cyber defense services.
Authors
Jina Kang, Jeong-Han Yun, Hyesong Choi
Abstract
Security operations centers are beginning to delegate detection, triage, and containment to Autonomous Cyber Defense (ACD) agents at execution time. Erroneous high-impact playbooks can disrupt critical services and create accountability gaps among deployers, SOC supervisors, vendors, managed service providers, and infrastructure providers. We propose SecODD, an execution-time governance framework for ACD that adapts the Operational Design Domain concept from autonomous-vehicle regulation to cyber autonomy. It defines the authorized scope of automated defensive actions, models when meaningful human intervention must remain possible, and specifies auditable due-care conditions through a Mitigation Safe Harbor. Four engineering requirements---tamper-evident logging, human-in-the-loop orchestration, rollback playbooks, and safe-state transition controls---are evaluated through a per-scenario audit matrix that illustrates how two ACD failure modes can be translated into pass/fail audit outcomes. The result is an evidence-oriented framework that authorizes ACD only when autonomous actions are bounded, interruptible, recoverable, and reconstructable from evidence.