An Ensemble Detection Method for Shilling Attacks Based on Features of Automatic Extraction

Faced with the evolving attacks in recommender systems, many detection features have been proposed by human engineering and used in supervised or unsupervised detection methods. However, the detection features extracted by human engineering are usually aimed at some specific types of attacks. To further detect other new types of attacks, the traditional methods have to re-extract detection features with high knowledge cost. To address these limitations, the method for automatic extraction of robust features is proposed and then an Adaboost-based detection method is presented. Firstly, to obtain robust representation with prior knowledge, unlike uniform corruption rate in traditional mLDA(marginalized Linear Denoising Autoencoder), different corruption rates for items are calculated according to the ratings’ distribution. Secondly, the ratings sparsity is used to weight the mapping matrix to extract low-dimensional representation. Moreover, the uniform corruption rate is also set to the next layer in mSLDA(marginalized Stacked Linear Denoising Autoencoder) to extract the stable and robust user features. Finally, under the robust feature space, an Adaboost-based detection method is proposed to alleviate the imbalanced classification problem. Experimental results on the Netflix and Amazon review datasets indicate that the proposed method can effectively detect various attacks.

A Novel Shilling Attack Detection Model Based on Particle Filter and Gravitation

With the rapid development of e-commerce, the security issues of collaborative filtering recommender systems have been widely investigated. Malicious users can benefit from injecting a great quantities of fake profiles into recommender systems to manipulate recommendation results. As one of the most important attack methods in recommender systems, the shilling attack has been paid considerable attention, especially to its model and the way to detect it. Among them, the loose version of Group Shilling Attack Generation Algorithm (GSAGenl) has outstanding performance. It can be immune to some PCC (Pearson Correlation Coefficient)-based detectors due to the nature of anti-Pearson correlation. In order to overcome the vulnerabilities caused by GSAGenl, a gravitation-based detection model (GBDM) is presented, integrated with a sophisticated gravitational detector and a decider. And meanwhile two new basic attributes and a particle filter algorithm are used for tracking prediction. And then, whether an attack occurs can be judged according to the law of universal gravitation in decision-making. The detection performances of GBDM, HHT-SVM, UnRAP, AP-UnRAP Semi-SAD,SVM-TIA and PCA-P are compared and evaluated. And simulation results show the effectiveness and availability of GBDM.

Generating A New Shilling Attack for Recommendation Systems

A collaborative filtering-based recommendation system has been an integral part of e-commerce and e-servicing.To keep the recommendation systems reliable,authentic,and superior,the security of these systems is very crucial.Though the existing shilling attack detection methods in collaborative filtering are able to detect the standard attacks,in this paper,we prove that they fail to detect a new or unknown attack.We develop a new attack model,named Obscure attack,with unknown features and observed that it has been successful in biasing the overall top-N list of the target users as intended.The Obscure attack is able to push target items to the top-N list as well as remove the actual rated items from the list.Our proposed attack is more effective at a smaller number of k in top-k similar user as compared to other existing attacks.The effectivity of the proposed attack model is tested on the MovieLens dataset,where various classifiers like SVM,J48,random forest,and naïve Bayes are utilized.

DPIF:A Framework for Distinguishing Unintentional Quality Problems From Potential Shilling Attacks

Maliciously manufactured user profiles are often generated in batch for shilling attacks.These profiles may bring in a lot of quality problems but not worthy to be repaired.Since repairing data always be expensive,we need to scrutinize the data and pick out the data that really deserves to be repaired.In this paper,we focus on how to distinguish the unintentional data quality problems from the batch generated fake users for shilling attacks.A two-steps framework named DPIF is proposed for the distinguishment.Based on the framework,the metrics of homology and suspicious degree are proposed.The homology can be used to represent both the similarities of text and the data quality problems contained by different profiles.The suspicious degree can be used to identify potential attacks.The experiments on real-life data verified that the proposed framework and the corresponding metrics are effective.

Securing Recommender Systems Against Shilling Attacks Using Social-Based Clustering

Abstract Recommender systems （RS） have been found supportive and practical in e-commerce and been established as useful aiding services. Despite their great adoption in the user communities, RS are still vulnerable to unscrupulous producers who try to promote their products by shilling the systems. With the advent of social networks new sources of information have been made available which can potentially render RS more resistant to attacks. In this paper we explore the information provided in the form of social links with clustering for diminishing the impact of attacks. We propose two algorithms, CLUTR and WCLUTR, to combine clustering with ＂trust＂ among users. We demonstrate that CLuTR and WCLUTR enhance the robustness of RS by experimentally evaluating them on data from a public consumer recommender system Epinions.com.

A COMPARATIVE STUDY ON SHILLING DETECTION METHODS FOR TRUSTWORTHY RECOMMENDATIONS

Uncovering shilling attackers hidden in recommender systems is very crucial to enhance the robustness and trustworthiness of product recommendation. Many shilling attack detection algorithms have been proposed so far, and they exhibit complementary advantage and disadvantage towards various types of attackers. In this paper, we provide a thorough experimental comparison of several well-known detectors, including supervised C4.5 and NB, unsupervised PCA and MDS, semi-supervised HySAD methods, as well as statistical analysis methods. MovieLens 100K is the most widely-used dataset in the realm of shilling attack detection, and thus it is selected as the benchmark dataset. Meanwhile, seven types of shilling attacks generated by average-filling and random-filling model are compared in our experiments. As a result of our analysis, we show clearly causes and essential characteristics insider attackers that might determine the success or failure of different kinds of detectors.

在线拍卖中的Shill出价识别模型

在线拍卖这种新的交易方式已经开始逐渐深入到人们的日常生活,但消费者在享受在线拍卖自由与便捷的同时,也屡屡受到欺诈,这严重影响了交易的诚信,Shill出价就是在线拍卖中典型的欺诈形式之一。如果能够识别Shill出价,就可以有目标地跟踪交易过程,使得Shill出价免于发生或尽量减少发生。在不改变费用结构的前提下,本文运用关联规则,在充分考虑Shill出价特点的基础上,分析卖者和买者间的关系,试图为Shill出价的识别提供一种行之有效的方法。实验结果证明了该方法的有效性。

A Robust Collaborative Recommendation Algorithm Based on k-distance and Tukey M-estimator

The existing collaborative recommendation algorithms have lower robustness against shilling attacks.With this problem in mind,in this paper we propose a robust collaborative recommendation algorithm based on k-distance and Tukey M-estimator.Firstly,we propose a k-distancebased method to compute user suspicion degree(USD).The reliable neighbor model can be constructed through incorporating the user suspicion degree into user neighbor model.The influence of attack profiles on the recommendation results is reduced through adjusting similarities among users.Then,Tukey M-estimator is introduced to construct robust matrix factorization model,which can realize the robust estimation of user feature matrix and item feature matrix and reduce the influence of attack profiles on item feature matrix.Finally,a robust collaborative recommendation algorithm is devised by combining the reliable neighbor model and robust matrix factorization model.Experimental results show that the proposed algorithm outperforms the existing methods in terms of both recommendation accuracy and robustness.

从shill到toadeater

人们通常把假装购物以引诱顾客上当的卖假货者的同伙或勾引赌客入局的赌棍同伙称之为“托儿”。英语中的叫法是 shill 或 shillaber。早在中世纪前后外国就有“托儿”了。他们活动的场所往往是 marketfair(集市),最早的“托儿”是走江湖卖假药的江湖医生的“雇员”,他往往是善于辞令且有表演才能的人;许多世纪以前,蟾蜍(toad)

From similarity perspective: a robust collaborative filtering approach for service recommendations

Collaborative filtering (CF) is a technique commonly used for personalized recommendation and Web service quality-of-service (QoS) prediction. However, CF is vulnerable to shilling attackers who inject fake user profiles into the system. In this paper, we first present the shilling attack problem on CF-based QoS recommender systems for Web services. Then, a robust CF recommendation approach is proposed from a user similarity perspective to enhance the resistance of the recommender systems to the shilling attack. In the approach, the generally used similarity measures are analyzed, and the DegSim (the degree of similarities with top k neighbors) with those measures is selected for grouping and weighting the users. Then, the weights are used to calculate the service similarities/differences and predictions. We analyzed and evaluated our algorithms using WS-DREAM and Movielens datasets. The experimental results demonstrate that shilling attacks influence the prediction of QoS values, and our proposed features and algorithms achieve a higher degree of robustness against shilling attacks than the typical CF algorithms.

Robust Recommendation Algorithm Based on Kernel Principal Component Analysis and Fuzzy C-means Clustering

The existing recommendation algorithms have lower robustness in facing of shilling attacks. Considering this problem, we present a robust recommendation algorithm based on kernel principal component analysis and fuzzy c-means clustering. Firstly, we use kernel principal component analysis method to reduce the dimensionality of the original rating matrix, which can extract the effective features of users and items. Then, according to the dimension-reduced rating matrix and the high correlation characteristic between attack profiles, we use fuzzy c-means clustering method to cluster user profiles, which can realize the effective separation of genuine profiles and attack profiles. Finally, we construct an indicator function based on the attack detection results to decrease the influence of attack profiles on the recommendation, and incorporate it into the matrix factorization technology to design the corresponding robust recommendation algorithm. Experiment results indicate that the proposed algorithm is superior to the existing methods in both recommendation accuracy and robustness.

SOFT CONTROL ON COLLECTIVE BEHAVIOR OF A GROUP OF AUTONOMOUS AGENTS BY A SHILL AGENT

This paper asks a new question： how can we control the collective behavior of self-organized multi-agent systems？ We try to answer the question by proposing a new notion called ＇Soft Control＇ which keeps the local rule of the existing agents in the system. We show the feasibility of soft control by a case study. Consider the simple but typical distributed multi-agent model proposed by Vicsek et al. for flocking of birds： each agent moves with the same speed but with different headings which are updated using a local rule based on the average of its own heading and the headings of its neighbors. Most studies of this model are about the self-organized collective behavior, such as synchronization of headings. We want to intervene in the collective behavior （headings） of the group by soft control. A specified method is to add a special agent, called a ＇Shill＇, which can be controlled by us but is treated as an ordinary agent by other agents. We construct a control law for the shill so that it can synchronize the whole group to an objective heading. This control law is proved to be effective analytically and numerieally. Note that soft control is different from the approach of distributed control. It is a natural way to intervene in the distributed systems. It may bring out many interesting issues and challenges on the control of complex systems.