Giseop Noh

Robust Sybil attack defense with information level in online Recommender Systems

As the major function of Recommender Systems (RSs) is recommending commercial items to potential consumers (i.e., system users), providing correct information of RS is crucial to both RS providers and system users. The influence of RS over Online Social Networks (OSNs) is expanding rapidly, whereas malicious users continuously try to attack the RSs with fake identities (i.e., Sybils) by manipulating the information in the RS adversely. In this paper, we propose a novel robust recommendation algorithm called RobuRec which exploits a distinctive feature, admission control. RobuRec provides highly trusted recommendation results since RobuRec predicts appropriate recommendations regardless of whether the ratings are given by honest users or by Sybils thanks to the power of admission control. To demonstrate the performance of RobuRec, we have conducted extensive experiments with various datasets as well as diverse attack scenarios. The evaluation results confirm that RobuRec outperforms the comparable schemes such as PCA and LTSMF significantly in terms of Prediction Shift (PS) and Hit Ratio (HR).

PSD: Practical Sybil detection schemes using stickiness and persistence in online recommender systems

The main function of recommender systems (RSs) is to recommend user-customized information to customers or system users. Correct and useful information is crucial for both customers and service providers. The influence of RSs is expanding over the Internet. However, criminal users try to manipulate the results of RSs with fake identities (i.e., Sybils) for financial gain. Effective metrics are consequently required for defense against Sybil attack. In this paper, we first explore two metrics, stickiness and persistence, from the perspective of the RS security domain. We then propose practical detecting schemes, Dynamic Sybil Attack Monitoring on Recommender Systems (DySy-Rec) and Fuzzy rule-based DySy-Rec (FDySy-Rec), which apply stickiness and persistence in two real datasets from real movie RSs. To demonstrate the effectiveness and potential of DySy-Rec and FDySy-Rec, we conducted extensive experiments on the inclusion of more diverse and smart types of attacks. The experimental results show that the proposed schemes achieve substantial performance improvement compared with previous statistical approaches in terms of precision and recall. Finally, the results confirm the practical possibilities of exploiting stickiness and persistence in the fight against dynamic Sybil attacks in online RSs.

Toward trustworthy social network services: A robust design of recommender systems

In recent years, electronic commerce and online social networks (OSNs) have experienced fast growth, and as a result, recommendation systems (RSs) have become extremely common. Accuracy and robustness are important performance indexes that characterize customized information or suggestions provided by RSs. However, nefarious users may be present, and they can distort information within the RSs by creating fake identities (Sybils). Although prior research has attempted to mitigate the negative impact of Sybils, the presence of these fake identities remains an unsolved problem. In this paper, we introduce a new weighted link analysis and influence level for RSs resistant to Sybil attacks. Our approach is validated through simulations of a broad range of attacks, and it is found to outperform other state-of-the-art recommendation methods in terms of both accuracy and robustness.

RobuRec: Robust Sybil attack defense in online recommender systems

With the growth of Internet usage and online social networks, the online Recommender Systems are becoming popular among system users. Although the influence of the recommender systems is expanding, the possibility of residing fake identities (Sybils) from nefarious users increase due to various reasons. To mitigate the impact of such users, several approaches are proposed. However, the need for robust algorithms is still necessary regarding recommender systems since the small portion of Sybils can distort the accuracy of predictions extremely. We propose a novel robust recommendation algorithm (RobuRec) using information level and admission control. The performance of RobuRec is experimented on various recommendation datasets with all possible Sybil attacks. The evaluation result shows that RobuRec can improve prediction error by 21% and 49% compared to two comparable schemes (LTSMF [23] and PCA [24], respectively). On all datasets and against various attack strategies, in turn, our RobuRec scheme shows the best peformance in terms of prediction shift.

Follow spam detection based on cascaded social information

In the last decade we have witnessed the explosive growth of online social networking services (SNSs) such as Facebook, Twitter, RenRen and LinkedIn. While SNSs provide diverse benefits for example, forstering interpersonal relationships, community formations and news propagation, they also attracted uninvited nuiance. Spammers abuse SNSs as vehicles to spread spams rapidly and widely. Spams, unsolicited or inappropriate messages, significantly impair the credibility and reliability of services. Therefore, detecting spammers has become an urgent and critical issue in SNSs. This paper deals with Follow spam in Twitter. Instead of spreading annoying messages to the public, a spammer follows (subscribes to) legitimate users, and followed a legitimate user. Based on the assumption that the online relationships of spammers are different from those of legitimate users, we proposed classification schemes that detect follow spammers. Particularly, we focused on cascaded social relations and devised two schemes, TSP-Filtering and SS-Filtering, each of which utilizes Triad Significance Profile (TSP) and Social status (SS) in a two-hop subnetwork centered at each other. We also propose an emsemble technique, Cascaded-Filtering, that combine both TSP and SS properties. Our experiments on real Twitter datasets demonstrated that the proposed three approaches are very practical. The proposed schemes are scalable because instead of analyzing the whole network, they inspect user-centered two hop social networks. Our performance study showed that proposed methods yield significantly better performance than prior scheme in terms of true positives and false positives.

STA : Sybil Type-aware Robust Recommender System

With a rapid development of internet, many users these days refer to various recommender sites when buying items, movies, music and more. However, there are malicious users (Sybil) who raise or lower item ratings intentionally in these recommender sites. And as a result, a recommender system (RS) may recommend incomplete or inaccurate results to normal users. We suggest a recommender algorithm to separate ratings generated by users into normal ratings and outlier ratings, and to minimize the effects of malicious users. Specifically, our algorithm first ensures a stable RS against three kinds of attack models (Random attack, Average attack, and Bandwagon attack) which are the main recent security issues in RS. To prove the performance of the method of suggestion, we conducted performance analysis on real world data that we crawled. The performance analysis demonstrated that the suggested method performs well regardless of Sybil size and type when compared to existing algorithms.

Trustor clustering with an improved recommender system based on social relationships

Abstract As we face a deluge of information in the modern world, the importance of recommender systems (RSs) that recommend relevant items to users has increased. The majority of existing RS schemes observe the prior ratings history of consumers to identify preferred items. However, current RSs suffer from the cold start problem, and their performance is dismal when new users or items appear. In order to address the cold start problem, a new type of solution that exploits social network features has been proposed. Many such social RSs analyze trustor–trustee relationships to discover latent social features shared between trustor and trustee. Since social relationships between trustors and trustees are directed, but not reciprocal, it is not guaranteed that a trustee has features in common with its trustors. Moreover, existing schemes are based on the assumption of independence between trustors who follow the same trustee, and therefore fail to recognize quintessential factors shared by the trustors. We posit that trustors who follow the same trustee have features in common. Based on the assumption that trustors who endorse the same trustee share similar tastes, we propose a new latent feature called Matrix S, and develop two novel RS algorithms that learn these latent features. We conduct an extensive performance evaluation using large scale real-world datasets, and observe that our proposed methods are not only more accurate than existing schemes but also show potential extensibility.

Power users are not always powerful: The effect of social trust clusters in recommender systems

Abstract A recommender system (RS) is one that provides optimized information to users in an over-supply situation. The key to an RS is the accurate prediction of the behavior of the user. The matrix factorization (MF) method is used for this prediction in the early stages, and based on the recent development of social network service (SNS), social information is also utilized to improve the accuracy of prediction. In this paper, we use an RS internal trust cluster for the first time to further improve performance and analyze the characteristics of trust clusters. We propose a new approach, a trust-aware network (TAN) RS, to exploit these trust clusters. We also explore the impact and influence of power users in a social network-based RS using TAN RS, and analyze the impact of power users and clusters. From our experiments, we find that the use of TAN RS can enhance the prediction accuracy of RSs; we also show that power users and the sizes of clusters are not significant, and that normal users and ordinary sizes of clusters contribute to a reduction in prediction errors in social RSs.

State Information Based Recommendation Algorithm for Minimizing the Malicious User's Influence*

ABSTRACT With the extreme development of Internet, recently most users r efer the sites with the various Recommendation Systems (RSs) when they want to buy some stuff, movie and music. Howeve r, the possibilities of the Sybils with the malicious behaviors may exists in these RSs sites in which Sybils intentionally inc rease or decrease the rating values. The RSs cannot play an accurate role of the proper recommendations to the general norm al users. In this paper, we divide the given rating values into the stable or unstable states and propose a system information based recommendation algorithm that minimizes the malicious user’s influence. To evaluate the performance of the proposed s cheme, we directly crawl the real trace data from the famous movie site and analyze the performance. After that, we showed p roposed scheme performs well compared to existing algorithms. Keywords: Recommendation system, Sybil, Social network I.서 론 * 최근 소셜 네트워크 서비스가 활성화되면서 사용자들에게 적합한 아이템들을 추천해 줄 수 있는 추천

Analysis & visualization on movie's popularity and reviews

Information visualization with the movies trend of audiences and reviews becomes important. Movie makers are not only want to know their movies popularity based on the number of audiences but also check their movies evaluation from people who see the movies. If they hope that the movie is a box office hit, they should identify the correlations between audiences and reviews. Our visualization tools try to find the movies moods by using normalized graphs and showing the relationships. Experiments on real dataset from a movie site (naver-movie.com) and Korean Film Council containing the number of customers in each movie. We find some unique relations and discover the word-of-mouth effects from our results.