面向推荐应用的差分隐私方案综述

大数据时代背景下,各行各业希望能基于用户行为数据来训练推荐模型,为用户提供精准推荐,所用数据的共性特点为总量庞大、携带敏感信息、易于获取。推荐系统在带来精准推荐和市场盈利的同时也正在实时分享着用户的隐私数据,差分隐私保护技术作为一门隐私保护技术,能够巧妙地解决推荐应用中存在的隐私泄露问题,其优势在于不需要考虑攻击方所具备的任何相关的背景知识、严格地对隐私保护进行了定义、提供了量化评估方法来保证数据集(在不同参数条件下)所提供的隐私保护水平具有可比较性。首先简述了差分隐私的概念和主流推荐算法的近期研究成果,其次重点分析了差分隐私与推荐算法相结合的应用情况,涉及的推荐算法有矩阵分解、深度学习推荐、协同过滤等,并对基于差分隐私保护的推荐算法的准确性进行了对比实验;然后讨论了与每种推荐算法结合的使用场景以及目前仍存在的问题,最后对基于差分隐私的推荐算法的未来发展方向提出了有效建议。     

基于联邦学习和区块链技术的TAP规则处理系统

触发执行编程（Trigger-Action Programming,TAP）为用户联动物联网（Internet of Things,IoT）设备提供了便捷的编程范式。利用机器学习对用户已编辑的TAP规则进行分析,实现TAP规则推荐和生成等功能可以提升用户体验。但TAP规则可能包含个人隐私信息,用户对上传和分享TAP信息存在顾虑。文章提出了基于联邦学习和区块链技术的TAP规则处理系统,用户可在本地进行TAP模型训练,无需上传隐私数据。为解决集中式服务器单点故障和防范恶意模型参数上传的问题,文章利用区块链技术改进集中式TAP联邦学习架构。用户将本地模型更新的累积梯度传输给区块链中的矿工,进行异常识别和交叉验证。矿工委员会整合正常用户提供的累积梯度,得到的全局模型作为一个新区块的数据,链接到区块链上,供用户下载使用。文章采用轻量级无监督的非负矩阵分解方法验证了提出的基于联邦学习和区块链的分布式学习架构的有效性。实验证明该联邦学习架构能有效保护TAP数据中的隐私,并且区块链中的矿工能够很好地识别恶意模型参数,确保了模型的稳定性。     

联邦推荐系统的协同过滤冷启动解决方法

基于联邦学习的推荐系统可以在保护用户隐私的情况下,联合多方数据,提升推荐系统的性能,已经成为推荐领域的研究热点之一.联邦协同过滤是联邦推荐系统中最经典及最常用的算法之一.然而,针对联邦协同过滤系统的冷启动问题的研究工作相对较少.针对这一问题,本文提出了一种基于安全内积协议的解决方案.具体地,在系统中添加新用户或新物品时...     

基于云联邦的差分隐私保护动态推荐模型

为提高推荐系统的准确性和个性化水平,同时保护用户的隐私,文章提出一种基于云联邦的差分隐私保护动态推荐模型（P2RCF）。该模型采用注意力机制动态调整融合长短期用户兴趣,增强推荐系统的灵活性,同时引入差分隐私技术和云联邦技术保护用户的隐私信息。文章在公共数据集上进行了实验,实验结果表明,该模型可以在保护用户数据隐私的同时提高推荐的准确性和个性化水平。     

面向用户隐私保护的联邦安全树算法

针对联邦学习算法在用户行为预测中存在的准确率低和运行效率不高等问题,提出一种无损失的联邦学习安全树（FLSectree）算法。首先,通过对损失函数的推导,证明损失函数的一阶偏导数与二阶偏导数为敏感数据,采用特征索引序列的扫描和分裂来返回加密后的最佳分裂点,以保护敏感数据不被泄露;接着,通过对实例空间的更新来继续向下分裂并寻找下一个最佳分裂点,直至满足终止条件后结束训练;最后,利用训练后的结果使得各参与方得到本地算法参数。实验结果表明,FLSectree算法能够在保护数据隐私的前提下有效提高用户行为预测算法的准确率和训练效率,与联邦学习FATE（Federated AI Technology Enabler）框架中的SecureBoost算法相比,FLSectree算法在用户行为预测中的准确率提高了9.09%,运行时间降低了87.42%,训练结果与集中式Xgboost算法一致。     

面向用户隐私保护的联邦安全树算法

针对联邦学习算法在用户行为预测中存在的准确率低和运行效率不高等问题，提出一种无损失的联邦学习安全树（FLSectree）算法。首先，通过对损失函数的推导，证明损失函数的一阶偏导数与二阶偏导数为敏感数据，采用特征索引序列的扫描和分裂来返回加密后的最佳分裂点，以保护敏感数据不被泄露；接着，通过对实例空间的更新来继续向下分裂并寻找下一个最佳分裂点，直至满足终止条件后结束训练；最后，利用训练后的结果使得各参与方得到本地算法参数。实验结果表明，FLSectree算法能够在保护数据隐私的前提下有效提高用户行为预测算法的准确率和训练效率，与联邦学习FATE（Federated AI Technology Enabler）框架中的SecureBoost算法相比，FLSectree算法在用户行为预测中的准确率提高了9.09%，运行时间降低了87.42%，训练结果与集中式Xgboost算法一致。     

面向用户的支持用户掉线的联邦学习数据隐私保护方法

联邦学习是解决多组织协同训练问题的一种有效手段,但是现有的联邦学习存在不支持用户掉线、模型API泄露敏感信息等问题。文章提出一种面向用户的支持用户掉线的联邦学习数据隐私保护方法,可以在用户掉线和保护的模型参数下训练出一个差分隐私扰动模型。该方法利用联邦学习框架设计了基于深度学习的数据隐私保护模型,主要包含两个执行协议:服务器和用户执行协议。用户在本地训练一个深度模型,在本地模型参数上添加差分隐私扰动,在聚合的参数上添加掉线用户的噪声和,使得联邦学习过程满足(ε,δ)-差分隐私。实验表明,当用户数为50、ε=1时,可以在模型隐私性与可用性之间达到平衡。     

基于跨域关联与隐私保护的深度推荐模型

推荐系统能够根据用户的偏好有效地过滤信息,已被广泛应用于各行各业,但随着用户数量的爆炸式增长,数据稀疏性和冷启动问题日益严重.多源数据融合可以有效缓解数据稀疏和冷启动情况下的推荐精度,其主要思想是融合用户在其他方面的辅助信息来填补缺失值,以优化目标服务的推荐准确度,受到了研究者的青睐,但由于数据之间的关联引入了更为严重的隐私泄露风险.针对以上问题,提出一种基于跨域关联与隐私保护的深度推荐模型,设计一种具有多源数据融合和差分隐私保护特征的深度学习协同推荐方法.该方法一方面融合辅助领域信息以提高推荐的精确度,同时修正异常点的偏差,改善推荐系统的性能;另一方面针对数据融合中的数据安全问题,基于差分隐私模型在协同训练过程中加入噪音以保证数据的安全性.为了更好地评价推荐系统中的长尾效应,首次提出一种新的评价指标-发现度,用以度量推荐算法发现用户隐性需求的能力.基于已有算法进行了性能对比与分析,实验结果证明,所提方法在保证隐私安全的前提下,比现有方法具有更好的推荐精度和多样性,能够有效地发现用户的隐性需求.     

面向隐私保护联邦学习的医学影像目标检测算法

联邦学习技术是一种新型多机构协同训练模型范式,广泛应用于多领域,其中模型参数隐私保护是一个关键问题.针对CT影像综合性病灶检测任务,提出隐私保护的联邦学习算法.首先部署松散耦合的客户端-服务器架构;其次在各客户端使用改进的RetinaNet检测器,引入上下文卷积和后向注意力机制;最后完成联邦训练.各客户端使用局部更新策略,采用自适应训练周期,局部目标函数中加入了限制项;服务器使用自适应梯度裁剪策略和高斯噪声差分隐私算法更新全局模型参数.在DeepLesion数据集上的消融分析说明了算法各部分的重要性.实验结果表明,改进的RetinaNet检测器有效地提升了多尺度病灶的检测精度.与集中数据训练模型范式相比,联邦学习所得模型性能略低(mAP分别为75.33％和72.80％),但训练用时缩短近38％,有效地实现了隐私保护、通信效率和模型性能的良好权衡.     

联邦学习安全防御与隐私保护技术研究

联邦学习(federated learning, FL)在多个参与方不直接进行数据传输的前提下共同完成模型训练,充分发挥各方数据价值;然而,由于联邦学习的固有缺陷以及存储和通信的安全问题,其在实际应用场景中仍面临多种安全与隐私威胁。首先阐述了FL面临的安全攻击和隐私攻击;然后针对这两类典型攻击分别总结了最新的安全防御机制和隐私保护手段,包括投毒攻击防御、后门攻击防御、搭便车攻击防御、女巫攻击防御以及基于安全计算与差分隐私的防御手段。通过对联邦学习的现有风险和相应防御手段的系统梳理,展望了联邦学习未来的研究挑战与发展方向。     

知识驱动的推荐系统：现状与展望

个性化推荐系统能够根据用户的个性化偏好和需要,自动、快速、精准地为用户提供其所需的互联网资源,已成为当今大数据时代应用最广泛的信息检索系统,具有巨大的商业应用价值。近年来,随着互联网海量数据的激增,人工智能技术的快速发展与普及,以知识图谱为代表的大数据知识工程日益受到学界和业界的高度关注,也有力地推动推荐系统和个性化推荐技术也迈入到知识驱动与赋能的发展阶段。将知识图谱中蕴含的丰富知识作为有用的辅助信息引入推荐系统,不仅能够有效应对数据稀疏、语义失配等传统推荐系统难以避免的问题,还能帮助推荐系统产生多样化、可解释的推荐结果,并更好地完成跨领域推荐、序列化推荐等具有挑战性的推荐任务,从而提升各类实际推荐场景中的用户满意度。本文将现有融入知识图谱的各种推荐模型按其采用的推荐算法与面向的推荐场景不同进行分类,构建科学、合理的分类体系。其中,按照推荐方法的不同,划分出基于特征表示的和基于图结构的两大类推荐模型;按推荐场景划分,特别关注多样化推荐、可解释推荐、序列化推荐与跨领域推荐。然后,我们在各类推荐模型中分别选取代表性的研究工作进行介绍,还简要对比了各个模型的特点与优劣。此外,本文还结合当下人工智能技术和应用的发展趋势,展望了认知智能推荐系统的发展前景,具体包括融合多模态知识的推荐系统,具有常识理解能力的推荐系统,以及解说式、劝说式、抗辩式推荐系统。本文的综述内容和展望可作为推荐系统未来研究方向的有益参考。     

一种改进的系统间隐私保持协同过滤推荐算法

针对系统间协同过滤推荐过程中的隐私泄露问题,以RSA公钥密码系统和安全多方计算SMC理论为基础,提出一个安全计算模型SCM,将安全计算模型SCM应用到系统间协同过滤中,得到一个有效的隐私保持协同过滤推荐算法。算法利用安全矢量积计算用户的相似度,防止了第三方的恶意串通。实验表明,该算法不但可以保护用户的隐私不泄露给协同合作的系统,而且提高了推荐算法的精度,特别是对用户数据稀疏的小站点。     

A privacy-preserving recommendation method with clustering and locality-sensitive hashing

Nowadays, there is a significant increase in information, resulting in information overload. Recommendation systems have been widely adopted, and they can help users find information relevant to their interests. However, a malicious attacker can infer users' private information via recommendations. To solve problems of data sparseness, enormous high-dimensional data, the cold start problem and privacy protection in an intelligent recommender system, this study proposes a privacy-preserving collaborative filtering recommendation method with clustering and locality-sensitive hashing. First, we cluster users according to their characteristic information to obtain sub-rating matrices. We use the latent factor model to predict and fill in the missing ratings in those matrices. Second, we combine the sub-rating matrices into a complete rating matrix, subsequently, we obtained the neighbors of the target user by analyzing the similarity of the users. We use a locality-sensitive hashing algorithm to reduce the dimensionality of the user rating data and build an index that could quickly obtain the neighbors of the target user. Finally, we predict the target user's ratings and provide recommendations to the target user. Through experiments, our study shows that our method can deal with the problems of data sparseness and cold start problems well and the accuracy of the intelligent recommendation system has been improved. In addition, we use hash techniques to search for the neighbors, which effectively protects the privacy of the user.     

Product recommendation with temporal dynamics

In many E-commerce recommender systems, a special class of recommendation involves recommending items to users in a life cycle. For example, customers who have babies will shop on Diapers.com within a relatively long period, and purchase different products for babies within different growth stages. Traditional recommendation algorithms produce recommendation lists similar to items that the target user has accessed before (content filtering), or compute recommendation by analyzing the items purchased by the users who are similar to the target user (collaborative filtering). Such recommendation paradigms cannot effectively resolve the situation with a life cycle, i.e., the need of customers within different stages might vary significantly. In this paper, we model users’ behavior with life cycles by employing hand-crafted item taxonomies, of which the background knowledge can be tailored for the computation of personalized recommendation. In particular, our method first formalizes a user’s long-term behavior using the item taxonomy, and then identifies the exact stage of the user. By incorporating collaborative filtering into recommendation, we can easily provide a personalized item list to the user through other similar users within the same stage. An empirical evaluation conducted on a purchasing data collection obtained from Diapers.com demonstrates the efficacy of our proposed method.     

Progress in context-aware recommender systems — An overview

Recommender Systems are the set of tools and techniques to provide useful recommendations and suggestions to the users to help them in the decision-making process for choosing the right products or services. The recommender systems tailored to leverage contextual information (such as location, time, companion or such) in the recommendation process are called context-aware recommender systems. This paper presents a review on the continual development of context-aware recommender systems by analyzing different kinds of contexts without limiting to any specific application domain. First, an in-depth analysis is conducted on different recommendation algorithms used in context-aware recommender systems. Then this information is used to find out that how these techniques deals with the curse of dimensionality, which is an inherent issue in such systems. Since contexts are primarily based on users’ activity patterns that leads to the development of personalized recommendation services for the users. Thus, this paper also presents a review on how this contextual information is represented (either explicitly or implicitly) in the recommendation process. We also presented a list of datasets and evaluation metrics used in the setting of CARS. We tried to highlight that how algorithmic approaches used in CARS differ from those of conventional RS. In that, we presented what modification or additions are being applied on the top of conventional recommendation approaches to produce context-aware recommendations. Finally, the outstanding challenges and research opportunities are presented in front of the research community for analysis     

Improving neighborhood based Collaborative Filtering via integrated folksonomy information

Personalized recommender systems which can provide people with suggestions according to individual interests usually rely on Collaborative Filtering (CF). The neighborhood based model (NBM) is a common choice when implementing such recommenders due to the intuitive nature; however, the recommendation accuracy is a major concern. Current NBM based recommenders mostly address the accuracy issue based on the rating data alone, whereas research on hybrid recommender systems suggests that users enjoy specifying feedback about items across multiple dimensions. In this work we aim to improve the accuracy of NBM via integrating the folksonomy information. To achieve this objective, we first propose the folksonomy network (FN) to analyze the item relevance described by the folksonomy data. We subsequently integrate the obtained folksonomy information into the global-optimization based NBM for making multi-source based recommendations. Experiments on the MovieLens dataset suggest positive results, which prove the efficiency of our strategy.     

联邦学习中的隐私保护技术研究综述

近年来,联邦学习成为解决机器学习中数据孤岛与隐私泄露问题的新思路。联邦学习架构不需要多方共享数据资源,只要参与方在本地数据上训练局部模型,并周期性地将参数上传至服务器来更新全局模型,就可以获得在大规模全局数据上建立的机器学习模型。联邦学习架构具有数据隐私保护的特质,是未来大规模数据机器学习的新方案。然而,该架构的参数交互方式可能导致数据隐私泄露。目前,研究如何加强联邦学习架构中的隐私保护机制已经成为新的热点。从联邦学习中存在的隐私泄露问题出发,探讨了联邦学习中的攻击模型与敏感信息泄露途径,并重点综述了联邦学习中的几类隐私保护技术：以差分隐私为基础的隐私保护技术、以同态加密为基础的隐私保护技术、以安全多方计算（SMC）为基础的隐私保护技术。最后,探讨了联邦学习中隐私保护中的若干关键问题,并展望了未来研究方向。     

Collaborative user modeling with user-generated tags for social recommender systems

With the popularity of social media services, the sheer amount of content is increasing exponentially on the Social Web that leads to attract considerable attention to recommender systems. Recommender systems provide users with recommendations of items suited to their needs. To provide proper recommendations to users, recommender systems require an accurate user model that can reflect a user’s characteristics, preferences and needs. In this study, by leveraging user-generated tags as preference indicators, we propose a new collaborative approach to user modeling that can be exploited to recommender systems. Our approach first discovers relevant and irrelevant topics for users, and then enriches an individual user model with collaboration from other similar users. In order to evaluate the performance of our model, we compare experimental results with a user model based on collaborative filtering approaches and a vector space model. The experimental results have shown the proposed model provides a better representation in user interests and achieves better recommendation results in terms of accuracy and ranking.     

A survey of autoencoder-based recommender systems

In the past decade,recommender systems have been widely used to provide users with personalized products and services.However,most traditional recommender systems are still facing a challenge in dealing with the huge volume,complexity,and dynamics of information.To tackle this challenge,many studies have been conducted to improve recommender system by integrating deep learning techniques.As an unsupervised deep learning method,autoencoder has been widely used for its excellent performance in data dimensionality reduction,feature extraction,and data reconstruction.Meanwhile,recent researches have shown the high efficiency of autoencoder in information retrieval and recommendation tasks.Applying autoencoder on recommender systems would improve the quality of recommendations due to its better understanding of users,demands and characteristics of items.This paper reviews the recent researches on autoencoder-based recommender systems.The differences between autoencoder-based recommender systems and traditional recommender systems are presented in this paper.At last,some potential research directions of autoencoder-based recommender systems are discussed.