From resilience to antifragility: embracing a new era in cybersecurity (www.techradar.com)

Radware’s CTO argues that generative AI has shifted cybersecurity from a resilience problem to an antifragility imperative: instead of merely restoring systems after attacks, defenses must learn and improve from each exposure. GenAI is changing the battleground—moving threats into the application layer and spawning AI hacking agents, indirect prompt injection (IPI) zero‑click attacks, and a shift from an API‑centric to an agent‑driven economy. While defenders already use AI to cut mean time to resolution from hours to minutes, the pace and unpredictability of AI‑led threats make reactive resilience insufficient. Practically, antifragile security means embedding autonomous intelligence (AI agents in SOCs that anticipate and act), protecting “inference perimeters” (model endpoints) via visibility, versioning and behavioral profiling, and treating red‑teaming as continual training for self‑healing architectures. Radware points to real incidents where teams replaced static defenses with adaptive defense agents to counter AI‑generated JavaScript attacks—illustrating the shift from recovery to reinvention. The piece urges a probabilistic/stochastic defense posture grounded in adaptive models so systems not only withstand disruption but improve because of it, making antifragility a strategic requirement for future cyber leadership.