To address the issues of path blockage and crash triggering difficulties in directed fuzzing (DF) caused by symbol stripping and complex state logic in Internet of things (IoT) firmware, a program state-aware fuzzing method, StateFuzz, is proposed. Firstly, static taint analysis is utilized to extract critical slices related to vulnerabilities. Subsequently, a call-graph-based hierarchical semantic recovery technique is adopted to reconstruct semantics within the slices, and large language models (LLM) are leveraged to accurately identify two categories of critical variables which are path constraints and vulnerability triggers. Finally, a “distance-state” dual-modal collaborative guidance mechanism is constructed. By integrating runtime monitoring with dynamic feedback information, entropy-guided energy scheduling and directed mutation of test cases are executed. Experimental results demonstrate that StateFuzz improves vulnerability discovery capability by 25% compared to that of the state-of-the-art baseline tool AntDFuzz, increases execution efficiency by 50.5%, and successfully discovers 11 undisclosed 0-day vulnerabilities. Ablation experiments further verify the critical role of each module in enhancing vulnerability mining performance, demonstrating the ability to efficiently break through state logic and precisely trigger vulnerabilities.
本文利用大语言模型(Large Language Model, LLM)的深度代码理解能力[19],基于反编译伪代码[20]识别关键变量以增强状态感知。针对符号剥离导致LLM推理偏差[21],造成语义缺失及状态感知困难问题,提出StateFuzz方法。首先,通过静态污点分析提取关键切片;随后,利用基于调用图的层级语义恢复技术,以库函数为锚点自底向上重构上下文语义,辅助LLM精准识别路径约束与漏洞触发变量;通过运行时监控变量熵值与分支敏感度动态校准误报;最后,构建“距离—状态”双模态协同制导机制,执行差异化能量调度与定向变异,以突破逻辑阻塞并精准触发漏洞。
AL-GARADI MALI, MOHAMEDA, AL-ALIA K, et al. A survey of machine and deep learning methods for Internet of Things (IoT) security[J]. IEEE Communications Surveys & Tutorials, 2020,22(3):1646-1685.
[2]
KHANS, MAZHART, SHAHZADT, et al. Antenna systems for IoT applications: a review[J]. Discover Sustainability, 2024,5(1):No.412.
[3]
COSTINA, ZADDACHJ, FRANCILLONA, et al. A large-scale analysis of the security of embedded firmwares[C]//Proceedings of the 23rd USENIX Conference on Security Symposium. Berkeley, USA: USENIX Association, 2014:95-110.
BÖHMEM, PHAMV T, NGUYENM D, et al. Directed greybox fuzzing[C]//Proceedings of the 2017 ACM SIGS AC Conference on Computer and Communications Security. New York, USA: ACM, 2017:2329-2344.
[7]
CHENH X, XUEY X, LIY K, et al. Hawkeye: towards a desired directed grey-box fuzzer[C]//Proceedings of the 2018 ACM SIGSAC Conference on Computer and Communications Security. New York, USA: ACM, 2018:2095-2108.
[8]
ZHENGY W, DAVANIANA, YINH, et al. FIRM-AFL: high-throughput greybox fuzzing of IoT firmware via augmented process emulation[C]//Proceedings of the 28th USENIX Conference on Security Symposium.Berkeley, USA: USENIX Association, 2019:1099-1114.
[9]
KIMM, KIMD, KIME, et al. FirmAE: towards large-scale emulation of IoT firmware for dynamic analysis[C]//Proceedings of the 36 th Annual Computer Security Applications Conference. New York, USA: ACM, 2020:733-745.
[10]
CHENJ Y, DIAOW R, ZHAOQ C, et al. IoTFuzzer: discovering memory corruptions in IoT through app-based fuzzing[C]//Proceedings 2018 Network and Distributed System Security Symposium. San Diego, USA: Internet Society, 2018. DOI:10.14722/ndss.2018.23159 .
[11]
FENGX T, SUNR X, ZHUX G, et al. Snipuzz: black-box fuzzing of IoT firmware via message snippet inference[C]//Proceedings of the 2021 ACM SIGSAC Conference on Computer and Communications Security. New York, USA: ACM, 2021:337-350.
[12]
QINC, PENGJ Q, LIUP Z, et al. UCRF: static analyzing firmware to generate under-constrained seed for fuzzing SOHO router[J]. Computers & Security, 2023,128:No.103157.
[13]
CUIX W, WANGY C, WEIQ. Sensitive target-guided directed fuzzing for IoT web services[J]. Computers, Materials & Continua, 2025,83(3):4939-4959.
[14]
REDININ, MACHIRYA, WANGR Y, et al. Karonte: detecting insecure multi-binary interactions in embedded firmware[C]//Proceedings of the 2020 IEEE Symposium on Security and Privacy. Piscataway, USA: IEEE, 2020:1544-1561.
[15]
MIFTAHS S, SRIVASTAVAA, KIMH, et al. SymbFuzz: symbolic execution guided hardware fuzzing[C]//Proceedings of the 58th IEEE/ACM International Symposium on Microarchitecture. New York, USA: ACM, 2025:1477-1490.
[16]
WANGY H, JIAX K, LIUY W, et al. Not all coverage measurements are equal: fuzzing by coverage accounting for input prioritization[C]//Proceedings of the 2020 Network and Distributed System Security Symposium. San Diego, USA: Internet Society, 2020. DOI:10.14722/ndss.2020.24422 .
[17]
CHENX, CUIN N, PANZ, et al. Critical variable state-aware directed greybox fuzzing[C]//Proceedings of the 2025 IEEE/ACM 47th International Conference on Software Engineering. Piscataway, USA: IEEE, 2025:141-152.
[18]
SEREBRYANYK, BRUENINGD, POTAPENKOA, et al.AddressSanitizer: a fast address sanity checker[C]//Proceedings of the 2012 USEIX Conference on Annual Technical Conference. Berkeley, USA: USENIX Association, 2012:No.12.
[19]
BROWNT B, MANNB, RYDERN, et al. Language models are few-shot learners[DB/OL]. (2020-05-28)[2026-03-18].
[20]
CHENQ B, LACOMISJ, SCHWARTZE J, et al. Augmenting decompiler output with learned variable names and types[DB/OL]. (2021-08-13)[2026-03-18].
[21]
HEJ X, IVANOVP, TSANKOVP, et al. Debin: predicting debug information in stripped binaries[C]//Proceedings of the 2018 ACM SIGSAC Conference on Computer and Communications Security. New York, USA: ACM, 2018:1667-1680.