CyberEvolver: Structured Self-Evolution for Cybersecurity Agents On the Fly

Fuente: arXiv
Enregistré dans:
Détails bibliographiques
Auteurs principaux: Fan, Yihe, Li, Changyi, Xu, Lichen, Pan, Xudong, Dai, Jiarun, Geng, Hong, Yang, Min
Format: Preprint
Publié: 2026
Sujets:
Accès en ligne:
Tags: Ajouter un tag
Pas de tags, Soyez le premier à ajouter un tag!
_version_ 1866913162360520704
author Fan, Yihe
Li, Changyi
Xu, Lichen
Pan, Xudong
Dai, Jiarun
Geng, Hong
Yang, Min
author_facet Fan, Yihe
Li, Changyi
Xu, Lichen
Pan, Xudong
Dai, Jiarun
Geng, Hong
Yang, Min
contents LLM-based agents are increasingly used for cybersecurity tasks, but most existing systems rely on fixed, human-designed scaffolds that struggle to adapt across diverse targets and failure modes. We introduce \textsc{CyberEvolver}, a self-evolving cybersecurity agent framework that iteratively revises its own scaffold based on experience from failed execution attempts. Self-evolution in cybersecurity is challenging because the space of possible scaffold changes is largely unstructured, execution feedback is sparse and often obscured by the environment, and low-diversity updates can cause errors to compound over repeated iterations. \textsc{CyberEvolver} addresses these challenges with a four-layer evolvable agent architecture that decomposes scaffold optimization into structured components, a trace-to-diagnosis mechanism that converts noisy execution logs into actionable revision signals, and a population-based beam search strategy that preserves diverse agent variants during evolution. We evaluate \textsc{CyberEvolver} on CTF challenges, vulnerability exploitation, and penetration-testing tasks using four open-source LLMs. Across these settings, \textsc{CyberEvolver} improves the seed agent's success rate by $13.6$\,\% on average, and outperforms six human-designed cybersecurity agents as well as two self-improvement methods adapted from other domains. These results suggest that scaffold self-evolution is a promising direction for building adaptive LLM agents for security testing.
format Preprint
id arxiv_https___arxiv_org_abs_2605_26195
institution arXiv
publishDate 2026
record_format arxiv
spellingShingle CyberEvolver: Structured Self-Evolution for Cybersecurity Agents On the Fly
Fan, Yihe
Li, Changyi
Xu, Lichen
Pan, Xudong
Dai, Jiarun
Geng, Hong
Yang, Min
Cryptography and Security
Artificial Intelligence
LLM-based agents are increasingly used for cybersecurity tasks, but most existing systems rely on fixed, human-designed scaffolds that struggle to adapt across diverse targets and failure modes. We introduce \textsc{CyberEvolver}, a self-evolving cybersecurity agent framework that iteratively revises its own scaffold based on experience from failed execution attempts. Self-evolution in cybersecurity is challenging because the space of possible scaffold changes is largely unstructured, execution feedback is sparse and often obscured by the environment, and low-diversity updates can cause errors to compound over repeated iterations. \textsc{CyberEvolver} addresses these challenges with a four-layer evolvable agent architecture that decomposes scaffold optimization into structured components, a trace-to-diagnosis mechanism that converts noisy execution logs into actionable revision signals, and a population-based beam search strategy that preserves diverse agent variants during evolution. We evaluate \textsc{CyberEvolver} on CTF challenges, vulnerability exploitation, and penetration-testing tasks using four open-source LLMs. Across these settings, \textsc{CyberEvolver} improves the seed agent's success rate by $13.6$\,\% on average, and outperforms six human-designed cybersecurity agents as well as two self-improvement methods adapted from other domains. These results suggest that scaffold self-evolution is a promising direction for building adaptive LLM agents for security testing.
title CyberEvolver: Structured Self-Evolution for Cybersecurity Agents On the Fly
topic Cryptography and Security
Artificial Intelligence
url https://arxiv.org/abs/2605.26195