Toward trustworthy recommender systems: An analysis of attack models and algorithm robustness

By Bamshad Mobasher, Robin D. Burke, Runa Bhaumik, Chad Williams. 2007

513 citations recorded by the source corpus.

Subfields: Information Systems, Management Science and Operations Research, Sociology and Political Science

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

In Information Systems

In Management Science and Operations Research

In Sociology and Political Science

Citations in the corpus

Poisoning Attacks to Graph-Based Recommender Systems (2018, methodology): “• Weighted Degree of Agreement (WDA) [21]: is feature is simply the numerator of the RDMA feature, i.”
Stable and Fair Classification (2019, background): “The dataset, that incorporates collected personal information, may be noisy due to measurement errors, privacy issues, or even data poisoning attacks [46, 56, 59, 6] and, hence, it is desirable that the fair classiﬁer also be stable against perturbations in the dataset.”
Stability of matrix factorization for collaborative filtering (2012, background): “Targeted Attacks include average attack (Lam & Riedl, 2004), segment attack and bandwagon attack (Mobasher et al., 2007).”
On the stability of recommendation algorithms (2010, background): “In particular, since recommender systems depend heavily on input from users, they are subject to manipulations and attacks [13-17].”
Recommender systems: from algorithms to user experience (2012, background): “Among others, Mobasher and his colleagues have produced a stream of research refining the understanding of the performance of a variety of attack methods against a variety of recommender algorithms (Sandvig et al. 2008; Mobasher et al. 2007; Williams et al. 2007).”
Recommender Systems (2016, unknown): “In recent years, a significant amount of effort has been devoted to the design of trustworthy recommender systems [444].”
Comparison of collaborative filtering algorithms (2011, methodology): “Several techniques affecting both neighbor-based [Mobasher et al. 2007] and model-based [Sandvig et al.”
Fast Differentially Private Matrix Factorization (2015, result): “This intuition is confirmed in the observed cache miss rates reported in the experiments in Section 6.”
Robustness of collaborative recommendation based on association rule mining (2007, background): “The random attack is a basic attack type that assigns random ratings to filler items, distributed around the global rating mean [9, 10].”
Music Recommendation (2016, background): “Recent works such as [27] have shown that for example nearest-neighbors algorithms can be vulnerable to various types of attacks whereas model-based approaches are often more robust in that respect.”
Evaluating Recommendation Systems (2011, background): “Such attempts to influence the recommendation are typically called attacks [43, 35].”
Privacy-preserving matrix factorization (2013, background): “Nevertheless, earlier studies [49, 37, 47, 1, 46] have identified multiple ways in which recommenders can abuse such information or expose the user to privacy threats.”
Shilling attacks against recommender systems: a comprehensive survey (2012, methodology): “Therefore, another class of attribute is introduced focusing on statistics across profiles called intraprofile detection attributes (Mobasher et al. 2007b).”
Collaborative Filtering Recommender Systems (2011, background): “Further, it is difficult to reliably detect the presence of malicious user profiles using standard recommender accuracy metrics, although classifer-based methods can detect some errant profiles [102].”