RSPapers | 基于隐私保护的推荐系统论文合集

嘿，记得给“机器学习与推荐算法”添加星标

近年来，推荐系统已经成为许多社交/购物/新闻平台中必不可少的组件。一方面，推荐系统为了更好的捕捉和建模用户的行为习惯以及历史偏好，需要大量收集用户和物品的属性信息以及二者的交互记录。另一方面，大量的用户行为记录以及用户私有属性信息虽然使得模型能够掌握用户的行为模式，但也不可避免的造成了用户敏感信息以及隐私问题的担忧。所以如何在保证用户隐私前提下挖掘数据价值是目前大数据背景下值得研究的课题。

早在推荐算法被提出来的初期，就一直有关于基于隐私保护的推荐系统的研究。比如，正如我们所熟知的Netflix大赛把研究人员关于推荐系统的研究热情带到了高点，但后来却因开放出来的数据集导致用户隐私泄露而叫停。而攻击的方法也很简单，文献[Arvind et al. 2008]通过将释放出的Netflix数据集与IMDb数据集进行关联就挖掘出了一部分用户的敏感信息，因此如何在提供推荐服务的同时保护用户的隐私问题变得越来越被人们重视。

所以，我们在增加了工业级的推荐系统、对话推荐系统等若干部分之后，决定将Privacy&Security RS部分增加到RSPapers项目中，希望能够开阔大家对于推荐系统领域的视野，以及能够为后续更好的科研或者付诸于实际产品有所帮助。目前该项目已经包含了15个小部分，累计获得3.2k星标，9位贡献者参与其中，期待对大家入门相关领域有所帮助。

通过此次调研，相关领域主要包含对于数据的隐私保护方法在推荐中的尝试，比如匿名化、差分隐私、本地化的差分隐私、同态加密算法、安全多方计算、联邦学习方法等与推荐方法的结合；以及机器学习思想在推荐中的尝试，比如对抗机器学习、对抗样本生成等。

接下来是推荐系统领域中隐私和安全问题相关的具体文献，希望大家可以从中发现更有价值的想法和问题。

Privacy&Security in RS

• John et al. Collaborative filtering with privacy via factor analysis. SIGIR, 2002.
• Arvind et al. Robust De-anonymization of Large Sparse Datasets. S&P, 2008.
• Udi et al. BlurMe: inferring and obfuscating user gender based on ratings. RecSys, 2012.
• Hua et al. Differentially Private Matrix Factorization. IJCAI, 2015.
• Erez et al. Secure Multi-Party Protocols for Item-Based Collaborative Filtering. Recsys, 2017.
• Meng et al. Personalized Privacy-Preserving Social Recommendation. AAAI, 2018.
• Shin et al. Privacy Enhanced Matrix Factorization for Recommendation with Local Differential Privacy. TKDE, 2018.
• Chen et al. Privacy Preserving Point-of-interest Recommendation Using Decentralized Matrix Factorization. AAAI, 2018.
• Istvan et al. Decentralized Recommendation Based on Matrix Factorization-A Comparison of Gossip and Federated Learning. PKDD, 2019.
• Aidmar et al. Efficient Privacy-Preserving Recommendations based on Social Graphs. RecSys, 2019.
• Muhammad et al. Federated Collaborative Filtering for Privacy-Preserving Personalized Recommendation System. arXiv, 2019.
• Chen et al. Attacking Recommender Systems with Augmented User Profiles. CIKM, 2020.
• Li et al. Federated Recommendation System via Differential Privacy. ISIT, 2020.
• Lin et al. FedRec: Federated Recommendation with Explicit Feedback. IEEE IS, 2020.
• Qi et al. FedRec: Privacy-Preserving News Recommendation with Federated Learning. arXiv, 2020.
• Zhang et al. GCN-Based User Representation Learning for Unifying Robust Recommendation and Fraudster Identification. SIGIR, 2020.
• Wang et al. Global and Local Differential Privacy for Collaborative Bandits. RecSys, 2020.
• Fang et al. Influence Function based Data Poisoning Attacks to Top-N Recommender Systems. WWW, 2020.
• Lin et al. Meta Matrix Factorization for Federated Rating Predictions. SIGIR, 2020.
• Zhang et al. Practical Data Poisoning Attack against Next-Item Recommendation. arXiv, 2020.
• Beigi et al. Privacy-Aware Recommendation with Private-Attribute Protection using Adversarial Learning. WSDM, 2020.
• Hu et al. PrivNet: Safeguarding Private Attributes in Transfer Learning for Recommendation. EMNLP, 2020.
• Chai et al. Secure Federated Matrix Factorization. IEEE IS, 2020.
• Chen et al. Secure Social Recommendation based on Secret Sharing. ECAI, 2020.
• Cohen et al. A Black-Box Attack Model for Visually-Aware Recommenders. WSDM, 2021.
• Huang et al. Data Poisoning Attacks to Deep Learning Based Recommender Systems. NDSS, 2021.
• Wu et al. FedGNN: Federated Graph Neural Network for Privacy-Preserving Recommendation. arXiv, 2021.
• Zhang et al. Graph Embedding for Recommendation against Attribute Inference Attacks. arXiv, 2021.