paperSeptember 2026Unreviewed

PrimSynth: An Agentic Approach to Discover, Validate, and Synthesize Exploit Primitives for Linux Kernel Vulnerabilities

Pengfei Wang, Anying Chen, Danjun Liu, Xu Zhou, Wei Xie

Abstract

Linux kernel vulnerabilities are critical to downstream systems. Despite extensive research on automated kernel exploitation, a fundamental challenge remains the conceptual gap between abstract exploit strategies and concrete technical operations. To fill this gap, this paper introduces a systematic characterization that formalizes six classes of exploit primitives from logical capability to validatable effect. Then, an extended exploit strategy representation is proposed, which couples primitiv

Cite

@misc{wang2026primsynth,
  title = {{PrimSynth: An Agentic Approach to Discover, Validate, and Synthesize Exploit Primitives for Linux Kernel Vulnerabilities}},
  author = {Pengfei Wang and Anying Chen and Danjun Liu and Xu Zhou and Wei Xie},
  year = {2026},
  month = sep,
  eprint = {2609.02647},
  archivePrefix = {arXiv},
  url = {https://arxiv.org/abs/2609.02647}
}