多视角特征融合的托攻击检测方法

doi:10.19678/j.issn.1000-3428.0070375

摘要/Abstract

摘要： 随着互联网平台的快速发展,推荐系统在提供个性化服务的同时,依旧存在由托攻击行为产生的推荐准确度下降问题。现有的托攻击检测算法大多以一个或几个评价指标从用户评分差异的视角进行评估,且较少考虑用户选择项目的偏好相关性,导致对用户行为特征建模不够全面,易造成检测误判率较高或适用攻击模式较少。针对这一问题,提出一种多视角特征融合的托攻击检测(MVFF-SAD)算法。根据变分自编码器(VAE)在获取用户评分潜在特征与分布特征的基础上,从评分长短期依赖、概率密度分布的视角学习用户概貌的时空分布特征,结合用户历史偏好相关性,利用神经网络模型进行多视角特征融合,形成具有综合检测能力的用户概貌表示,以提升托攻击检测精度。实验结果表明,所提算法对虚假用户的检测精度有较大提升,在大多数情况下均能达到95%以上。在不同填充率和不同攻击规模下,该算法对托攻击行为均具有较好的检测效果,且具有较好的鲁棒性。

关键词: 推荐系统, 托攻击, 用户偏好, 用户概貌, 神经网络

Abstract: As Internet platforms progress rapidly, recommendation systems are challenged by degraded accuracy due to shilling attacks, while still providing personalized services. Existing shilling attack detection algorithms mostly focus on a single or a few evaluation metrics from the perspective of user rating differences and seldom consider the preference correlation of the items selected by users. Consequently, user behavior characteristics are modeled insufficiently, resulting in high misdetection rates or limited applicability to different attack patterns. To address this issue, this paper proposes a Multi-View Feature Fusion Shilling Attack Detection (MVFF-SAD) algorithm. Based on the latent features and distribution characteristics of user ratings obtained through a Variational AutoEncoder (VAE), the algorithm learns the spatiotemporal distribution characteristics of user profiles from the perspectives of short- and long-term dependencies of ratings and probability density distributions. By combining the historical preference correlation of users, the algorithm uses a neural network model for multi-view feature fusion to form a comprehensive user profile representation with enhanced detection capability, thereby improving shilling attack detection accuracy. Experimental results show that the proposed algorithm significantly improves the detection accuracy for fake users, achieving an accuracy of over 95% in most cases. The proposed algorithm also demonstrates good detection performance under different filling rates and attack scales, and exhibits strong robustness.

Key words: recommendation system, shilling attack, user preference, user profile, neural networks

中图分类号:

TP391

贾俊杰, 李天乐, 刘世龙. 多视角特征融合的托攻击检测方法[J]. 计算机工程, 2026, 52(7): 302-312.

JIA Junjie, LI Tianle, LIU Shilong. Multi-View Feature Fusion Method for Shilling Attack Detection[J]. Computer Engineering, 2026, 52(7): 302-312.

https://www.ecice06.com/CN/Y2026/V52/I7/302

参考文献

[1] KHALIL H A. Towards optimizing hybrid movie recommender systems[J]. Revue d’Intelligence Artificielle, 2024, 38(1):159-173.
[2] AHMADIAN YAZDI H, SEYYED MAHDAVI S J, AHMADIAN YAZDI H. Dynamic educational recommender system based on improved LSTM neural network[J]. Scientific Reports, 2024, 14(1): 4381.
[3] SINGH P K, PRAMANIK P K D, SINHABABU N, et al. Detecting unknown shilling attacks in recommendation systems[J]. Wireless Personal Communications, 2024, 137: 259-286.
[4] 雷梦宁, 丁爱玲, 王新美, 等. 基于混合特征值的托攻击检测算法[J]. 计算机技术与发展, 2021, 31(10): 87-92. LEI M N, DING A L, WANG X M, et al. Shilling attack detection algorithm based on hybrid eigenvalue[J]. Computer Technology and Development, 2021, 31(10): 87-92. (in Chinese)
[5] CHIRITA P A, NEJDL W, ZAMFIR C. Preventing shilling attacks in online recommender systems[C]//Proceedings of the 7th Annual ACM International Workshop on Web Information and Data Management. New York, USA: ACM Press, 2005: 67-74.
[6] BURKE R, MOBASHER B, WILLIAMS C, et al. Classification features for attack detection in collaborative recommender systems[C]//Proceedings of the 12th ACM SIGKDD International Conference on Knowledge Discovery and Data Mining. New York, USA: ACM Press, 2006: 542-547.
[7] WILLIAMS C A, MOBASHER B, BURKE R. Defending recommender systems: detection of profile injection attacks[J]. Service Oriented Computing and Applications, 2007, 1(3): 157-170.
[8] BARBIERI J, ALVIM L G M, BRAIDA F, et al. Simulating real profiles for shilling attacks: a generative approach[J]. Knowledge-Based Systems, 2021, 230: 107390.
[9] RANI S, KAUR M, KUMAR M, et al. Detection of shilling attack in recommender system for YouTube video statistics using machine learning techniques[J]. Soft Computing, 2023, 28:945.
[10] CAO J, WU Z A, MAO B, et al. Shilling attack detection utilizing semi-supervised learning method for collaborative recommender system[J]. World Wide Web, 2013, 16: 729-748.
[11] 叶榕, 邵剑飞, 邵建龙. 基于用户长短期偏好的个性化推荐[J]. 吉林大学学报(理学版), 2024, 62(3): 615-628. YE R, SHAO J F, SHAO J L. Personalized recommendation based on users’ long-term and short-term preferences[J]. Journal of Jilin University (Science Edition), 2024,62(3): 615-628.(in Chinese)
[12] 雒晓辉, 吴云, 王晨星, 等. 基于用户长短期偏好的序列推荐模型[J]. 计算机科学, 2023, 50(4): 47-55. LUO X H, WU Y, WANG C X, et al. Sequential recommendation model based on user’s long and short term preference[J]. Computer Science, 2023, 50(4): 47-55. (in Chinese)
[13] 吕成戍. 基于用户项目属性偏好的协同过滤推荐算法[J]. 计算机技术与发展, 2018, 28(4): 152-156, 160. Lü C S. A robust collaborative filtering recommendation algorithm based on user preference of item attributes[J]. Computer Technology and Development, 2018, 28(4): 152-156,160. (in Chinese)
[14] 田俊峰, 蔡红云. 托攻击与推荐系统安全[J]. 河北大学学报(自然科学版), 2018, 38(6): 640-647,655. TIAN J F, CAI H Y. Shilling attacks and security of recommender systems[J]. Journal of Hebei University (Natural Science Edition), 2018, 38(6): 640-647,655. (in Chinese)
[15] LEE J S, ZHU D. Shilling attack detection—a new approach for a trustworthy recommender system[J]. INFORMS Journal on Computing, 2012, 24(1): 117-131.
[16] 卫星君, 顾清华. 针对协同过滤推荐系统的混淆托攻击模型[J]. 计算机与数字工程, 2018, 46(8): 1575-1579, 1696. WEI X J, GU Q H. Shilling attack model of obfuscation for collaborative filtering recommender system[J]. Computer[WT《Times New Roman》]& Digital Engineering, 2018, 46(8): 1575-1579,1696. (in Chinese)
[17] GAO M, YUAN Q, LING B, et al. Detection of abnormal item based on time intervals for recommender systems[J]. The Scientific World Journal, 2014, 2014(1): 845897.
[18] GUNES I. A novel clustered-based detection method for shilling attack in private environments[J]. PeerJ Computer Science, 2024, 10: e2137.
[19] 邵晨. 面向协同过滤推荐系统的托攻击检测方法研究[D]. 哈尔滨:哈尔滨理工大学,2023. SHAO C. Research on shilling attack detection for collaborative filtering recommender systems[D]. Harbin: Harbin University of Science and technology, 2023. (in Chinese)
[20] KARTHIKEYAN P, SELVI S T, NEERAJA G, et al. Prevention of shilling attack in recommender systems using discrete wavelet transform and support vector machine[C]//Proceedings of the 8th International Conference on Advanced Computing. Washington D.C., USA: IEEE Press, 2017: 99-104.
[21] 景慧昀, 周川, 贺欣. 针对人脸检测对抗攻击风险的安全测评方法[J]. 计算机科学, 2021, 48(7): 17-24. JING H Y, ZHOU C, HE X. Security evaluation method for risk of adversarial attack on face detection[J]. Computer Science, 2021, 48(7): 17-24. (in Chinese)
[22] ZHANG F Z, ZHANG Z N, ZHANG P, et al. UD-HMM: an unsupervised method for shilling attack detection based on hidden Markov model and hierarchical clustering[J]. Knowledge-Based Systems, 2018, 148: 146-166.
[23] CAI H Y, ZHANG F Z. An unsupervised approach for detecting group shilling attacks in recommender systems based on topological potential and group behaviour features[J]. Security and Communication Networks, 2021, 2021(1): 2907691.
[24] 吕成戍. 基于非对称半监督集成SVM的托攻击检测方法[J]. 运筹与管理, 2018, 27(8): 84. Lü C S. Shilling attack detection approach based on asymmetric semi-supervised ensemble SVM[J]. Operations Research and Management Science, 2018, 27(8): 84-91. (in Chinese)
[25] 卫星君, 顾清华. 基于特征指标推荐系统托攻击半监督检测[J]. 计算机应用研究, 2018, 35(7):2185-2188. WEI X J, GU Q H. Semi-supervised detection of shilling attack for recommender system based on characteristic index[J]. Application Research of Computers, 2018, 35(7): 2185-2188. (in Chinese)
[26] ZHOU Q Q, WU J X, DUAN L L. Recommendation attack detection based on deep learning[J]. Journal of Information Security and Applications, 2020, 52: 102493.
[27] 于金霞, 李佳昕, 李星宇, 等. 多尺度特征融合的托攻击检测方法[J]. 重庆邮电大学学报(自然科学版), 2023, 35(5):863-872. YU J X, LI J X, LI X Y, et al. Shilling attack detection method based on multi-scale feature fusion[J]. Journal of Chongqing University of Posts and Telecommunications(Natural Science Edition), 2023, 35(5):863-872.(in Chinese)
[28] SHAO C, ZHONG Y, SUN Y. Shilling attack detection for collaborative recommender systems: a gradient boosting method[J]. Mathematical Biosciences and Engineering, 2022, 19(7): 7248-7271.
[29] LI H, GAO M, ZHOU F T, et al. Fusing hypergraph spectral features for shilling attack detection[J]. Journal of Information Security and Applications, 2021, 63: 103051.
[30] SHAMS S, LEITH D. Attack detection using item vector shift in matrix factorisation recommenders[J]. ACM Transactions on Privacy and Security, 2023, 28(2):1-12.
[31] NARAYANAN P, VIVEKANANDAN K. Hybrid CNN and RNN-based shilling attack framework in social recommender networks[J]. EAI Endorsed Transactions on Scalable Information Systems, 2022, 9(35): e6.
[32] 雷梦宁, 丁爱玲, 王新美, 等. 基于混合特征值的托攻击检测算法[J]. 计算机技术与发展, 2021, 31(10):87-92. LEI M N, DING A L, WANG X M, et al. Shilling attack detection algorithm based on hybrid eigenvalue[J]. Computer Technology and Development, 2021, 31(10):87-92.(in Chinese)
[33] 王心浩. 基于变分自编码和监督原型网络的推荐系统托攻击检测方法研究[D]. 南京: 南京财经大学, 2023. WANG X H. Research on shilling attack detection method in recommender systems based on variational autoencoder and supervised prototype network[D]. Nanjing: Nanjing University of Finance and Economics,2023.
[34] LI L, WANG Z Q, LI C, et al. Collaborative filtering recommendation using fusing criteria against shilling attacks[J]. Connection Science, 2022, 34(1): 1678-1696.
[35] 方霖枫, 周仁杰. 基于图神经网络的挖掘潜在偏好图推荐算法[J]. 软件导刊, 2024, 23(10): 129-138. FANG L F, ZHOU R J. Recommendation algorithm for mining potential preference graph based on graph neural network[J]. Software Guide, 2024, 23(10): 129-138. (in Chinese)
[36] 李文涛, 高旻, 李华, 等. 一种基于流行度分类特征的托攻击检测算法[J]. 自动化学报, 2015, 41(9): 1563-1576. LI W T, GAO M, LI H, et al. An shilling attack detection algorithm based on popularity degree features[J]. Acta Automatica Sinica, 2015, 41(9): 1563-1576. (in Chinese)

选择文件类型/文献管理软件名称

选择包含的内容