SysEvolve: An AI-native, safe, autonomous adversarial attack-defense co-evolutionary system

2026-08-14

AI Summary

This study addresses the stagnation in cybersecurity attack-defense evolution and asymmetric AI capabilities by proposing a co-evolutionary framework. We construct an autonomous adversarial closed-loop system integrating simulation environments, LLM-based agents, and explainable defense mechanisms. Through three-tier end-to-end self-driven exercises, the system facilitates the co-evolution of AI-native security capabilities. Experimental results demonstrate that this approach increases attack success rates by over 25% and improves defense precision by three orders of magnitude, while successfully detecting real-world APT attacks. These findings effectively validate the feasibility and efficacy of autonomous adversarial engagement in driving security evolution.

Abstract

The rapid advancement of large language models (LLMs) has created a growing asymmetry in cybersecurity, where attack accelerates toward autonomous execution while defense remains predominantly human-intensive. Despite substantial prior work across cyber ranges, AI-driven attack, and AI-driven defense, this asymmetry persists. We trace it to a deeper root cause, that evolution itself has stalled on both sides at three layers. To overcome this, we propose co-evolution as the integrating insight, where attack and defense AI agents autonomously and safely drive each other's evolution through adversarial confrontation. Based on this insight, we present \sysevolve, comprising three co-designed components, \sysfield, \sysspear, and \sysarmor. \sysfield constructs realistic multi-host ranges. \sysspear generates efficient, safe attack schemes. \sysarmor performs real-time, interpretable defense. Together they form a self-driven adversarial loop restoring evolution at all three layers. In evaluation, \sysfield achieves zero-loss collection at 2.1\% overhead and orchestrates 257 CVEs into 1,148 ranges, \sysspear improves attack success by over 25\% over baseline LLMs, and \sysarmor achieves 10--1000$\times$ greater precision than prior systems and detects real APT attacks in production at Huawei and Sangfor. Our evaluation also reveals three findings about LLM agent capabilities. First, multi-step composition and larger topologies expose agent capability gaps hidden by single-step evaluations. Second, the bottleneck lies after initial access in post-compromise state utilization. Third, LLM agents are susceptible to environmental interference. When decoy endpoints are deployed in the range, agent timeouts triple and downstream completion disappears despite the success rates of initial accesses are unchanged.

Problem

• Cybersecurity asymmetry
• Adversarial co-evolution
• Autonomous attack-defense
• LLM agents
• Evolutionary stagnation

Innovation

• co-evolutionary system
• adversarial attack-defense
• AI-native cybersecurity
• autonomous agents
• interpretable defense