Robust multivariate autoregression for anomaly detection in dynamic product ratings

Günnemann, Nikou; Günnemann, Stephan; Faloutsos, Christos

doi:10.1145/2566486.2568008

Cited by 41 publications

(32 citation statements)

References 19 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…However, calculating the DegSim consumed a lot of time. Recently, Gnnemann et al [5] proposed a detection technique by analyzing temporal rating distributions. In addition, Gnnemann et al [6] presented a new detection approach based on the sound Bayesian framework to detect concerned anomalous ratings.…”

Section: Related Workmentioning

confidence: 99%

“…Collaborative filtering recommender systems (CFRSs) have proved to be one of the most popular RSs. However, CFRSs are highly vulnerable to shilling attacks due to their openness, which are injected with chosen profiles of abnormal ratings in order to control recommendation results to their benefits or decrease the trustworthiness of recommendation [5][6]. Therefore, constructing an effective detection method is crucial to detect attackers and remove them from CFRSs.…”

Section: Introductionmentioning

confidence: 99%

See 1 more Smart Citation

Defending Suspected Ratings in Collaborative Filtering Recommender Systems: A Fast Detection Method

Yang¹

2016

Proceedings of the Fourth International Conference on Information Science and Cloud Computing — PoS(ISCC2015)

View full text Add to dashboard Cite

Collaborative filtering recommender systems (CFRSs) are key components of the well-known E-commerce websites such as Amazon, Yelp etc., to make personalized recommendations. In practice, CFRSs are highly vulnerable to "shilling" attacks. Detection methods based on such attacks have received much attention. However, their detection accuracy is not fully acceptable especially when the attack size or filler size is small. In this paper, we solve the following task: Given the rating dataset, how can we spot abnormal ratings of users as well as keeping reasonable time-consumption? We propose a fast and effective detection method to detect such attacks, which consists of two phases. We firstly capture all suspected users by employing a topk similarity method for calculating the similarity between users. Finally, we continue to filter out more genuine users by analysing target items as far as possible. In addition, extensive experiments demonstrate that the detection performance of our method outperforms benchmarked methods. It is noteworthy that the recently published attack, PIA (power item attack) including PIA-AS, PIA-ID and PIA-NR can be detected by our proposed method. 2015 18-19, December, ISCC

show abstract

Section: Related Workmentioning

confidence: 99%

Section: Introductionmentioning

confidence: 99%

Defending Suspected Ratings in Collaborative Filtering Recommender Systems: A Fast Detection Method

Yang¹

2016

Proceedings of the Fourth International Conference on Information Science and Cloud Computing — PoS(ISCC2015)

View full text Add to dashboard Cite

show abstract

“…So far, there exists only a single method which analyzes temporal rating data under the presence of anomalies [7]. A potential drawback of [7], however, is the necessary aggregation/binning of the data.…”

Section: Introductionmentioning

confidence: 99%

“…A potential drawback of [7], however, is the necessary aggregation/binning of the data. When using a coarse aggregation, the temporal effects of the data are not well captured (in the extreme, all data is a single bin).…”

Section: Introductionmentioning

confidence: 99%

See 1 more Smart Citation

Detecting anomalies in dynamic rating data

Günnemann

Faloutsos

2014

Proceedings of the 20th ACM SIGKDD International Conference on Knowledge Discovery and Data Mining

Self Cite

View full text Add to dashboard Cite

Rating data is ubiquitous on websites such as Amazon, TripAdvisor, or Yelp. Since ratings are not static but given at various points in time, a temporal analysis of rating data provides deeper insights into the evolution of a product's quality. In this work, we tackle the following question: Given the time stamped rating data for a product or service, how can we detect the general rating behavior of users as well as time intervals where the ratings behave anomalous?We propose a Bayesian model that represents the rating data as sequence of categorical mixture models. In contrast to existing methods, our method does not require any aggregation of the input but it operates on the original time stamped data. To capture the dynamic effects of the ratings, the categorical mixtures are temporally constrained: Anomalies can occur in specific time intervals only and the general rating behavior should evolve smoothly over time. Our method automatically determines the intervals where anomalies occur, and it captures the temporal effects of the general behavior by using a state space model on the natural parameters of the categorical distributions. For learning our model, we propose an efficient algorithm combining principles from variational inference and dynamic programming. In our experimental study we show the effectiveness of our method and we present interesting discoveries on multiple real world datasets.

show abstract

Detecting deceptive engagement in social media by temporal pattern analysis of user behaviors: a survey

Xiao

Bertino

2017

WIREs Data Min & Knowl

View full text Add to dashboard Cite

Deceptive engagement in social media, such as spamming, commenting, or rating with automatic scripts, spreading fabricated facts, seriously affects users’ trust on online services. Given the large volumes of information generated by users, effectively spotting users involved such deceptive engagement has become a challenging problem. Recent research has shown that techniques for analyzing temporal behavioral patterns are critical to address such problem. In this study, we survey recent advances in these techniques. We first summarize three approaches to model temporal information. Then, by using representative application examples, we discuss recent approaches with respect to their applications to real‐world large‐scale social media. With a focus on the temporal perspective, we then discuss advantages and challenges of each approach. WIREs Data Mining Knowl Discov 2017, 7:e1210. doi: 10.1002/widm.1210 This article is categorized under: Fundamental Concepts of Data and Knowledge > Knowledge Representation Fundamental Concepts of Data and Knowledge > Motivation and Emergence of Data Mining Technologies > Prediction

show abstract

Robust multivariate autoregression for anomaly detection in dynamic product ratings

Cited by 41 publications

References 19 publications

Defending Suspected Ratings in Collaborative Filtering Recommender Systems: A Fast Detection Method

Defending Suspected Ratings in Collaborative Filtering Recommender Systems: A Fast Detection Method

Detecting anomalies in dynamic rating data

Detecting deceptive engagement in social media by temporal pattern analysis of user behaviors: a survey

Contact Info

Product

Resources

About