Malicious Software Adversarial Defense Model Based on Feature Severity Ranking

doi:10.3969/j.issn.1671-1122.2024.04.013

Abstract

Abstract:

The application of deep learning models in the detection of Android malware can continuously improved the accuracy of detection. However, with the proposal of adversarial examples, these examples can easily evade detection by deep learning models, leading to questions about the detection capabilities of deep learning models. To counteract adversarial attacks on Android malicious software, current approaches often employ adversarial training for defense. This paper addressed the limitation of adversarial training in dealing with various types of adversarial examples and proposed the concept of feature maliciousness. Feature maliciousness involved ranking features based on their malicious nature, and this ranked feature set was utilized to construct a malicious software adversarial defense model with adversarial defense capabilities, termed the feature maliciousness processing (FMP) detector. This model extracted high-maliciousness features from the software under consideration, mitigating the problem of model misclassification caused by adversarial perturbations. On the open-source dataset DefenceDroid, the feature selection method employed by the FMP detector significantly enhances the detection rates for various types of adversarial examples compared to adversarial training and other feature selection methods. Under multiple adversarial example attacks, the FMP detector demonstrats the highest level of robust performance.

Key words: Android malicious software, adversarial example, feature selection, deep learning

CLC Number:

TP309

XU Zirong, GUO Yanping, YAN Qiao. Malicious Software Adversarial Defense Model Based on Feature Severity Ranking[J]. Netinfo Security, 2024, 24(4): 640-649.

Figures/Tables 8

References 35

[1]	CHRISTIAAN B, SANDEEP C, TAYLOR D, et al. McAfee Labs Threats Report: November 2020[EB/OL]. (2020-11-10)[2023-11-10]. https://www.trellix.com/enterprise/zh-cn/assets/reports/rp-quarterly-threats-nov-2020.pdf.
[2]	GOODFELLOW I J, SHLENS J, SZEGEDY C. Explaining and Harnessing Adversarial Examples[EB/OL]. (2014-12-20)[2023-11-10]. https://arxiv.org/abs/1412.6572.
[3]	FENG Yu, ANAND S, DILLIG I, et al. Apposcopy: Semantics-Based Detection of Android Malware through Static Analysis[C]// ACM. 22nd ACM SIGSOFT International Symposium on the Foundations of Software Engineering. New York: ACM, 2014: 576-587.
[4]	FENG Yu, BASTANI O, MARTINS R, et al. Automated Synthesis of Semantic Malware Signatures Using Maximum Satisfiability[EB/OL]. (2016-08-22)[2023-11-10]. https://arxiv.org/abs/1608.06254.
[5]	ARZT S, RASTHOFER S, FRITZ C, et al. FlowDroid: Precise Context, Flow, Field, Object-Sensitive and Lifecycle-Aware Taint Analysis for Android Apps[J]. ACM SIGPLAN Notices, 2014, 49(6): 259-269.
[6]	PANDITA R, XIAO Xusheng, YANG Wei, et al. WHYPER: Towards Automating Risk Assessment of Mobile Applications[C]// USENIX. 22nd USENIX Security Symposium (USENIX Security 13). Berkeley: USENIX, 2013: 527-542.
[7]	TEUFL P, FERK M, FITZEK A, et al. Malware Detection by Applying Knowledge Discovery Processes to Application Metadata on the Android Market (Google Play)[J]. Security & Communication Networks, 2016, 9(5): 389-419.
[8]	ZHANG Mu, DUAN Yue, YIN Heng, et al. Semantics-Aware Android Malware Classification Using Weighted Contextual API Dependency Graphs[C]// ACM. 2014 ACM SIGSAC Conference on Computer and Communications Security(CCS). New York: ACM, 2014: 1105-1116.
[9]	YUAN Zhenlong, LU Yongqiang, WANG Zhaoguo, et al. Droid-Sec: Deep Learning in Android Malware Detection[J]. ACM SIGCOMM Computer Communication Review, 2014, 44(4): 371-372.
[10]	PEYNIRCI G, EMINAĞAOĞLU M, KARABULUT K. Feature Selection for Malware Detection on the Android Platform Based on Differences of IDF Values[J]. Journal of Computer Science and Technology, 2020, 35(4): 946-962.
[11]	CHEN Sen, XUE Minhui, FAN Lingling, et al. Automated Poisoning Attacks and Defenses in Malware Detection Systems: An Adversarial Machine Learning Approach[J]. Computers & Security, 2018(73): 326-344.
[12]	LI Xiangjun, KONG Ke, XU Su, et al. Feature Selection-Based Android Malware Adversarial Sample Generation and Detection Method[J]. IET Information Security, 2021, 15(6): 401-416.
[13]	HU Weiwei, TAN Ying. Generating Adversarial Malware Examples for Black-Box Attacks Based on GAN[C]// Springer. Data Mining and Big Data-7th International Conference (DMBD 2022). Heidelberg: Springer, 2022: 409-423.
[14]	KAWAI M, OTA K, DONG Mianxing. Improved MalGAN: Avoiding Malware Detector by Leaning Cleanware Features[C]// IEEE. The 1st International Conference on Artificial Intelligence in Information and Communication (ICAIIC). New York: IEEE, 2019: 40-45.
[15]	LI Heng, ZHOU Shiyao, YUAN Wei, et al. Adversarial-Example Attacks Toward Android Malware Detection System[J]. IEEE Systems Journal, 2019, 14(1): 653-656.
[16]	ONWUZURIKE L, MARICONTI E, ANDRIOTIS P, et al. MaMaDroid: Detecting Android Malware by Building Markov Chains of Behavioral Models (Extended Version)[J]. ACM Transaction on Information and System Security, 2019, 22(2): 1-34.
[17]	CHEN Xiao, LI Chaoran, WANG Derui, et al. Android HIV: A Study of Repackaging Malware for Evading Machine-Learning Detection[J]. IEEE Transactions on Information Forensics and Security, 2019(15): 987-1001.
[18]	PAPERNOT N, MCDANIEL P, JHA S, et al. The Limitations of Deep Learning in Adversarial Settings[C]// IEEE. The 1st IEEE European Symposium on Security and Privacy (Euro S&P). New York: IEEE, 2016: 372-387.
[19]	CARLINI N, WAGNER D. Towards Evaluating the Robustness of Neural Networks[C]// IEEE. The 38th IEEE Symposium on Security and Privacy (SP). New York: IEEE, 2017: 39-57.
[20]	SZEGEDY C, ZAREMBA W, SUTSKEVER I, et al. Intriguing Properties of Neural Networks[EB/OL]. (2014-02-19)[2023-11-10]. https://arxiv.org/abs/1312.6199.
[21]	RATHORE H, SAHAY S K, NIKAM P, et al. Robust Android Malware Detection System against Adversarial Attacks Using Q-Learning[J]. Information Systems Frontiers, 2021, 23(4): 867-882.
[22]	RATHORE H, SAHAY S K, DHILLON J, et al. Designing Adversarial Attack and Defence for Robust Android Malware Detection Models[C]// IEEE. The 51st Annual IEEE/IFIP International Conference on Dependable Systems and Networks(DSN). New York: IEEE, 2021: 29-32.
[23]	TAHERI R, JAVIDAN R, POORANIAN Z. Adversarial Android Malware Detection for Mobile Multimedia Applications in IoT Environments[J]. Multimedia Tools and Applications, 2021, 80(11): 16713-16729.
[24]	CHEN Yiming, YANG C H, CHEN G C. Using Generative Adversarial Networks for Data Augmentation in Android Malware Detection[C]// IEEE. The 2021 IEEE Conference on Dependable and Secure Computing(DSC). New York: IEEE, 2021: 1-8.
[25]	LI Heng, ZHOU Shiyao, YUAN Wei, et al. Robust Android Malware Detection against Adversarial Example Attacks[C]// ACM. The Web Conference 2021 (WWW’21). New York: ACM, 2021: 3603-3612.
[26]	CHEN Lingwei, HOU Shifu, YE Yanfang. SecureDroid: Enhancing Security of Machine Learning-Based Detection against Adversarial Android Malware Attacks[C]// ACM. The 33rd Annual Computer Security Applications Conference (ACSAC’17). New York: ACM, 2017: 362-372.
[27]	CLINE W C, MANDAR D B, SIDDHARTH A B, et al. DefenseDroid: A Modern Approach to Android Malware Detection[EB/OL]. (2021-12-07)[2023-11-10]. https://github.com/DefenseDroid/DefenseDroid.
[28]	MAHDAVIFAR S, KADIR A F A, FATEMI R, et al. Dynamic Android Malware Category Classification Using Semi-Supervised Deep Learning[C]// IEEE. 2020 IEEE International Conference on Dependable, Autonomic and Secure Computing, International Conference on Pervasive Intelligence and Computing, International Conference on Cloud and Big Data Computing, International Conference on Cyber Science and Technology Congress (DASC/PiCom/CBDCom/CyberSciTech). New York: IEEE, 2020: 515-522.
[29]	MAHDAVIFAR S, ALHADIDI D, GHORBANI A. Effective and Efficient Hybrid Android Malware Classification Using Pseudo-Label Stacked Auto-Encoder[J]. Journal of Network and Systems Management, 2022, 30(1): 1-34.
[30]	WANG Haoyu, SI Junjun, LI Hao, et al. RmvDroid: Towards a Reliable Android Malware Dataset with App Metadata[C]// IEEE. 16th International Conference on Mining Software Repositories (MSR2019). New York: IEEE, 2019: 404-408.
[31]	ALLIX K, BISSYANDÉ T F, KLEIN J, et al. AndroZoo: Collecting Millions of Android Apps for the Research Community[C]// IEEE. 2016 IEEE/ACM 13th International Conference on Mining Software Repositories (MSR2016). New York: IEEE, 2016: 468-471.
[32]	MOOSAVI-DEZFOOLI S M, FAWZI A, FROSSARD P. Deepfool: A Simple and Accurate Method to Fool Deep Neural Networks[C]// IEEE. 2016 IEEE Conference on Computer Vision and Pattern Recognition. New York: IEEE, 2016: 2574-2582.
[33]	SU Jiawei, VARGAS D V, SAKURAI K. One Pixel Attack for Fooling Deep Neural Networks[J]. IEEE Transactions on Evolutionary Computation, 2019, 23(5): 828-841.
[34]	WILSON E B, HILFERTY M M. The Distribution of Chi-Square[J]. National Academy of Sciences of the United States of America, 1931, 17(12): 684-688.
[35]	HIEMSTRA D. A Probabilistic Justification for Using TF×IDF Term Weighting in Information Retrieval[J]. International Journal on Digital Libraries, 2000, 3(2): 131-139.

特征名	恶意频率	良性频率	恶意度
permission.WRITE_SETTINGS	55.77%	10.28%	0.4549
permission.SYSTEM_ALERT_WINDOW	59.12%	11.27%	0.4787
permission.CHANGE_WIFI_STATE	59.00%	9.14%	0.4986
permission.GET_TASKS	68.12%	13.62%	0.5449
permission.MOUNT_UNMOUNT_FILESYSTEMS	59.25%	3.00%	0.5625

神经网络模型/ 对抗样本类型	对抗训练前	对抗训练后
神经网络模型/ 对抗样本类型	对抗样本1	对抗样本1	对抗样本2
检测器1	2.07%	99.97%	15.03%
检测器2-FMP	75.29%	75.29%	97.93%

特征数量/个	阈值
特征数量/个	0.2	0.4	0.5	0.6	0.8
128	0.8807	0.9069	0.9027	0.9073	0.8753
256	0.9278	0.9446	0.9434	0.9439	0.9252
384	0.8885	0.9168	0.9160	0.9052	0.8887
512	0.6874	0.6688	0.6684	0.6698	0.6683