Reducing hubness: A cause of vulnerability in recommender systems

Kazuo Hara, Ikumi Suzuki, Kei Kobayashi, Kenji Fukumizu

Research output: Chapter in Book/Report/Conference proceedingConference contribution

1 Citation (Scopus)

Abstract

It is known that memory-based collaborative filtering systems are vulnerable to shilling attacks. In this paper, we demonstrate that hubness, which occurs in high dimensional data, is exploited by the attacks. Hence we explore methods for reducing hubness in user-response data to make these systems robust against attacks. Using the MovieLens dataset, we empirically show that the two methods for reducing hubness by transforming a similarity matrix|(i) centering and (ii) conversion to a commute time kernel|can thwart attacks without degrading the recommendation performance.

Original languageEnglish
Title of host publicationSIGIR 2015 - Proceedings of the 38th International ACM SIGIR Conference on Research and Development in Information Retrieval
PublisherAssociation for Computing Machinery, Inc
Pages815-818
Number of pages4
ISBN (Electronic)9781450336215
DOIs
Publication statusPublished - 2015 Aug 9
Externally publishedYes
Event38th International ACM SIGIR Conference on Research and Development in Information Retrieval, SIGIR 2015 - Santiago, Chile
Duration: 2015 Aug 92015 Aug 13

Other

Other38th International ACM SIGIR Conference on Research and Development in Information Retrieval, SIGIR 2015
CountryChile
CitySantiago
Period15/8/915/8/13

Fingerprint

Collaborative filtering
Recommender systems
Data storage equipment

Keywords

  • Collaborative filtering
  • Hubness
  • Shilling attack

ASJC Scopus subject areas

  • Information Systems
  • Software

Cite this

Hara, K., Suzuki, I., Kobayashi, K., & Fukumizu, K. (2015). Reducing hubness: A cause of vulnerability in recommender systems. In SIGIR 2015 - Proceedings of the 38th International ACM SIGIR Conference on Research and Development in Information Retrieval (pp. 815-818). Association for Computing Machinery, Inc. https://doi.org/10.1145/2766462.2767823

Reducing hubness : A cause of vulnerability in recommender systems. / Hara, Kazuo; Suzuki, Ikumi; Kobayashi, Kei; Fukumizu, Kenji.

SIGIR 2015 - Proceedings of the 38th International ACM SIGIR Conference on Research and Development in Information Retrieval. Association for Computing Machinery, Inc, 2015. p. 815-818.

Research output: Chapter in Book/Report/Conference proceedingConference contribution

Hara, K, Suzuki, I, Kobayashi, K & Fukumizu, K 2015, Reducing hubness: A cause of vulnerability in recommender systems. in SIGIR 2015 - Proceedings of the 38th International ACM SIGIR Conference on Research and Development in Information Retrieval. Association for Computing Machinery, Inc, pp. 815-818, 38th International ACM SIGIR Conference on Research and Development in Information Retrieval, SIGIR 2015, Santiago, Chile, 15/8/9. https://doi.org/10.1145/2766462.2767823
Hara K, Suzuki I, Kobayashi K, Fukumizu K. Reducing hubness: A cause of vulnerability in recommender systems. In SIGIR 2015 - Proceedings of the 38th International ACM SIGIR Conference on Research and Development in Information Retrieval. Association for Computing Machinery, Inc. 2015. p. 815-818 https://doi.org/10.1145/2766462.2767823
Hara, Kazuo ; Suzuki, Ikumi ; Kobayashi, Kei ; Fukumizu, Kenji. / Reducing hubness : A cause of vulnerability in recommender systems. SIGIR 2015 - Proceedings of the 38th International ACM SIGIR Conference on Research and Development in Information Retrieval. Association for Computing Machinery, Inc, 2015. pp. 815-818
@inproceedings{b42d0a014fac41f5b2d77b7de2c7feb4,
title = "Reducing hubness: A cause of vulnerability in recommender systems",
abstract = "It is known that memory-based collaborative filtering systems are vulnerable to shilling attacks. In this paper, we demonstrate that hubness, which occurs in high dimensional data, is exploited by the attacks. Hence we explore methods for reducing hubness in user-response data to make these systems robust against attacks. Using the MovieLens dataset, we empirically show that the two methods for reducing hubness by transforming a similarity matrix|(i) centering and (ii) conversion to a commute time kernel|can thwart attacks without degrading the recommendation performance.",
keywords = "Collaborative filtering, Hubness, Shilling attack",
author = "Kazuo Hara and Ikumi Suzuki and Kei Kobayashi and Kenji Fukumizu",
year = "2015",
month = "8",
day = "9",
doi = "10.1145/2766462.2767823",
language = "English",
pages = "815--818",
booktitle = "SIGIR 2015 - Proceedings of the 38th International ACM SIGIR Conference on Research and Development in Information Retrieval",
publisher = "Association for Computing Machinery, Inc",

}

TY - GEN

T1 - Reducing hubness

T2 - A cause of vulnerability in recommender systems

AU - Hara, Kazuo

AU - Suzuki, Ikumi

AU - Kobayashi, Kei

AU - Fukumizu, Kenji

PY - 2015/8/9

Y1 - 2015/8/9

N2 - It is known that memory-based collaborative filtering systems are vulnerable to shilling attacks. In this paper, we demonstrate that hubness, which occurs in high dimensional data, is exploited by the attacks. Hence we explore methods for reducing hubness in user-response data to make these systems robust against attacks. Using the MovieLens dataset, we empirically show that the two methods for reducing hubness by transforming a similarity matrix|(i) centering and (ii) conversion to a commute time kernel|can thwart attacks without degrading the recommendation performance.

AB - It is known that memory-based collaborative filtering systems are vulnerable to shilling attacks. In this paper, we demonstrate that hubness, which occurs in high dimensional data, is exploited by the attacks. Hence we explore methods for reducing hubness in user-response data to make these systems robust against attacks. Using the MovieLens dataset, we empirically show that the two methods for reducing hubness by transforming a similarity matrix|(i) centering and (ii) conversion to a commute time kernel|can thwart attacks without degrading the recommendation performance.

KW - Collaborative filtering

KW - Hubness

KW - Shilling attack

UR - http://www.scopus.com/inward/record.url?scp=84953791320&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=84953791320&partnerID=8YFLogxK

U2 - 10.1145/2766462.2767823

DO - 10.1145/2766462.2767823

M3 - Conference contribution

AN - SCOPUS:84953791320

SP - 815

EP - 818

BT - SIGIR 2015 - Proceedings of the 38th International ACM SIGIR Conference on Research and Development in Information Retrieval

PB - Association for Computing Machinery, Inc

ER -