자율무인체계 환경의 센서-제어 가상화 경로 취약점 분석 자동화 연구

초록

In unmanned systems, state inconsistencies within sensor-to-control virtualization paths pose critical safety risks, yet traditional fuzzingstruggles to detect these complex vulnerabilities. This study proposes an automated vulnerability analysis framework utilizing reducedtarget harnesses and state-aware fuzzing with assertion-based oracles. The proposed approach models core virtualization mechanisms—such as dynamic device attachment and asynchronous streaming—to induce abnormal state transitions. This enables the effective screeningof latent semantic errors that do not cause immediate system crashes. Experimental results on the Xen hypervisor successfully identifiedand reproduced critical structural faults, including invalid resource reuse, partial rollback failures, and resource leaks. Because these faultscan trigger control deadline misses and sensor disconnections, this research establishes a vital new direction for securing cyber-physicalsystems (CPS).

키워드