TY - GEN
T1 - CCE&D
T2 - 29th Australasian Conference on Information Security and Privacy, ACISP 2024
AU - Zhang, Yanqiu
AU - Yu, Xiao
AU - Liu, Jinzhao
AU - Zhang, Li
AU - Li, Yuanzhang
AU - Tan, Yuan
N1 - Publisher Copyright:
© The Author(s), under exclusive license to Springer Nature Singapore Pte Ltd. 2024.
PY - 2024
Y1 - 2024
N2 - The correct setting of software configuration items is essential for improving software stability and ensuring safe, reliable operation. By contrast, potential configuration errors can have serious negative effects on software operation and even cause catastrophic consequences. Compared to traditional software, autonomous driving systems involve large amounts of data acquisition, processing, and real-time decision-making, and thus have a higher degree of configurability, making them more susceptible to safety issues from configuration errors. Most previous work on configuration failure diagnosis for autonomous driving systems focused on passive diagnosis after failure occurrence, making it difficult to detect potential untriggered configuration failures during system operation. In this paper, we propose CCE&D, which automatically infers configuration constraints from source code, detect configuration failures prior to configuration-specific deployment, preventing their occurrence in autopilot systems. Experimental results show the constraint rules covers 75% of the platform’s total configuration item constraints with 98.9% accuracy. Meanwhile, the accuracy of configuration error detection reaches 96.39%, and the purpose of configuration fault prevention is achieved.
AB - The correct setting of software configuration items is essential for improving software stability and ensuring safe, reliable operation. By contrast, potential configuration errors can have serious negative effects on software operation and even cause catastrophic consequences. Compared to traditional software, autonomous driving systems involve large amounts of data acquisition, processing, and real-time decision-making, and thus have a higher degree of configurability, making them more susceptible to safety issues from configuration errors. Most previous work on configuration failure diagnosis for autonomous driving systems focused on passive diagnosis after failure occurrence, making it difficult to detect potential untriggered configuration failures during system operation. In this paper, we propose CCE&D, which automatically infers configuration constraints from source code, detect configuration failures prior to configuration-specific deployment, preventing their occurrence in autopilot systems. Experimental results show the constraint rules covers 75% of the platform’s total configuration item constraints with 98.9% accuracy. Meanwhile, the accuracy of configuration error detection reaches 96.39%, and the purpose of configuration fault prevention is achieved.
KW - Autonomous Driving Systems
KW - configuration failure
KW - constraint inference
KW - static analysis
UR - http://www.scopus.com/inward/record.url?scp=85200466780&partnerID=8YFLogxK
U2 - 10.1007/978-981-97-5101-3_16
DO - 10.1007/978-981-97-5101-3_16
M3 - Conference contribution
AN - SCOPUS:85200466780
SN - 9789819751006
T3 - Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)
SP - 288
EP - 302
BT - Information Security and Privacy - 29th Australasian Conference, ACISP 2024, Proceedings
A2 - Zhu, Tianqing
A2 - Li, Yannan
PB - Springer Science and Business Media Deutschland GmbH
Y2 - 15 July 2024 through 17 July 2024
ER -