Download: PDF.
“Collaborative Recommendation Vulnerability to Focused Bias Injection Attacks” by R. Burke, B. Mobasher, R. Bhaumik, and C. Williams. In Proceedings of the Workshop on Privacy and Security Aspects of Data Mining, (Held at ICDM'05, Houston, Texas), Nov. 2005.
Significant vulnerabilities have recently been identified in collaborative recommender systems. Attackers who cannot be readily distinguished from ordinary users may inject biased data in an attempt to force the system to “adapt” in a manner advantageous to them. Researchers have studied simple attack models and their impact on a system's population of users. In this paper, we examine attacks that concentrate on a targeted set of users with similar tastes, biasing the system's responses to these users. Not only are such attacks more pragmatically beneficial for the attacker (since a particular item can be pushed to those most likely to buy it), but as we show, such attacks are also highly effective against both user-based and item-based algorithms. As a result, an attacker can mount such a “segmented” attack with little knowledge of the specific system being targeted and with strong likelihood of success.
Keywords: profile injection attacks, collaborative filtering, recommender systems, segmented attacks, item-based
Download: PDF.
BibTeX entry:
@inproceedings{BMBW05b,
author = {R. Burke and B. Mobasher and R. Bhaumik and C. Williams},
title = {Collaborative Recommendation Vulnerability to Focused Bias
Injection Attacks},
booktitle = {Proceedings of the Workshop on Privacy and Security
Aspects of Data Mining},
address = {Held at ICDM'05, Houston, Texas},
month = nov,
year = {2005},
url = {http://www.cacs.louisiana.edu/~icdm05/}
}