Vulnerability Mining of Deep Learning Framework for Model Generation Guided by Reinforcement Learning

被引：0
|
作者：
Pan L. [1 ]
Liu L. [1 ]
Luo S. [1 ]
Zhang Z. [1 ]
机构：
[1] School of Information and Electronics, Beijing Institute of Technology, Beijing
来源：
Beijing Ligong Daxue Xuebao/Transaction of Beijing Institute of Technology | 2024年 / 44卷 / 05期
关键词：
deep learning framework; generative model; reinforcement learning; vulnerability mining;
D O I：
10.15918/j.tbit1001-0645.2023.137
中图分类号：
学科分类号：
摘要：
In the existing methods, the vulnerability mining is randomly generating the structural information of the model according to application model, generating easily a large number of low-quality test cases, and seriously affecting the efficiency and effect of vulnerability mining. To solve this problem, a vulnerability mining method of deep learning framework was proposed based on a guiding model generation method with reinforcement learning. Firstly, frame state information during model running was extracted, including Softmax distance and program execution results, etc. Then the extracted frame running state information was taken as a reward variable to guide the generation of model structure and hyper-parameters, so as to improve the quality and efficiency of test case generation. Experimental results show that this method can find more vulnerability of deep learning frameworks under the same number of generated test cases, possessing high practical value. © 2024 Beijing Institute of Technology. All rights reserved.
引用
收藏
页码：521 / 529
页数：8
相关论文
共 19 条
  • [11] PHAM H V, LUTELLIER T, QI W, Et al., CRADLE: cross-backend validation to detect and localize bugs in deep learning libraries, 2019 IEEE/ACM 41st International Conference on Software Engineering, pp. 1027-1038, (2019)
  • [12] WANG Z, YAN M, CHEN J, Et al., Deep learning library testing via effective model generation, Proceedings of the 28th ACM Joint Meeting on European Software Engineering Conference and Symposium on the Foundations of Software Engineering, pp. 788-799, (2020)
  • [13] GUO Q, XIE X, LI Y, Et al., Audee: Automated testing for deep learning frameworks, Proceedings of the 35th IEEE/ACM International Conference on Automated Software Engineering, pp. 486-498, (2020)
  • [14] LI M, CAO J, TIAN Y, Et al., COMET: Coverage-guided model generation for deep learning library testing[J], ACM Transactions on Software Engineering and Methodology, 32, 5, pp. 1-34, (2023)
  • [15] WATKINS C J C H, DAYAN P., Q-learning[J], Machine Learning, 8, pp. 279-292, (1992)
  • [16] BAKER B, GUPTA O, NAIK N, Et al., Designing neural network architectures using reinforcement learning
  • [17] LI H, LIU H, JI X, Et al., Cifar10-dvs: an event-stream dataset for object classification, Frontiers in Neuroscience, 11, (2017)
  • [18] BALDOMINOS A, SAEZ Y, ISASI P., A survey of handwritten character recognition with mnist and emnist, Applied Sciences, 9, 15, (2019)
  • [19] KAYED M, ANTER A, MOHAMED H., Classification of garments from fashion MNIST dataset using CNN LeNet-5 architecture, 2020 International Conference on Innovative Trends in Communication and Computer Engineering, pp. 238-243, (2020)
12