Marke: Jana Thiessmeier
Variante: Taschenbuch
Eigenschaften:
Cyber defense must become more adaptive—without becoming less governable.Modern organizations operate across cloud platforms, identity fabrics, SaaS ecosystems, APIs, remote endpoints, software supply chains, operational technology, and machine-to-machine trust relationships. Adversaries exploit not only vulnerabilities, but also the predictability of defensive systems and the latency between evidence, authority, and action.Polymorphic Agentic Cyberdefense: Toward Bounded Autonomous Adaptation in Dynamic Adversarial Environments introduces PACD as a rigorous framework for cyber defense systems that can observe, reason, adapt, and act under adversarial uncertainty while remaining bounded by mission intent, explicit authority, safety constraints, reversibility, human oversight, controlled execution, verification, and auditability.Inside the monographA formal definition of Polymorphic Agentic Cyberdefense and a taxonomy of six classes of governed defensive polymorphismA constrained, partially observable model for defensive decision-making under uncertaintyAn eleven-layer, vendor-neutral reference architecture spanning evidence and reasoning, authority and governance, and execution and assuranceA hybrid algorithmic framework for evidence fusion, belief updating, threat-hypothesis generation, constrained planning, policy-portfolio selection, multi-agent coordination, fallback, rollback, and recoveryA contextual autonomy model that separates recommendation, authorization, execution, verification, and learningA progressive evaluation program covering analytical verification, simulation, emulation, human-in-the-loop studies, shadow operation, and bounded pilots At the center of PACD is guarded agency. Humans define mission intent, protected functions, delegated authority, prohibited actions, escalation thresholds, and accountability. Agents may operate at machine speed inside that envelope, but they may not define their own mission, expand their own authority, bypass policy, or directly exercise unrestricted production privileges.The book develops practical architectural concepts for security leaders and technical teams, including the safety governor, controlled execution brokers, mission models, action ontologies, evidence fabrics, context-sensitive autonomy tiers, decision packages, safe degradation, and zero-trust boundaries for agents.Written for: cybersecurity and AI researchers, security architects, CISOs, SOC and incident-response leaders, critical-infrastructure defenders, defense and national-security practitioners, AI-assurance specialists, policymakers, standards professionals, and executives responsible for high-consequence digital systems.This first research monograph is theoretical, formal, architectural, algorithmic, governance-oriented, and methodological. It reports no empirical performance results and makes no claim that PACD has already demonstrated superior effectiveness, operational safety, generalization, or production readiness. Instead, it establishes a coherent and testable foundation for future implementations, reproducible technical reports, and independent evaluation.The governing question is precise: Can cyber defense change fast enough to reduce adversary freedom of action while remaining mission-aware, reversible, accountable, and under human control?
