Robust collaborative filtering

@inproceedings{Mehta2007RobustCF,
  title={Robust collaborative filtering},
  author={Bhaskar Mehta and Thomas Hofmann and Wolfgang Nejdl},
  booktitle={RecSys '07},
  year={2007}
}
The widespread deployment of recommender systems has lead to user feedback of varying quality. While some users faithfully express their true opinion, many provide noisy ratings which can be detrimental to the quality of the generated recommendations. The presence of noise can violate modeling assumptions and may thus lead to instabilities in estimation and prediction. Even worse, malicious users can deliberately insert attack profiles in an attempt to bias the recommender system to their… 

Figures and Tables from this paper

Attack resistant collaborative filtering
TLDR
A new collaborative algorithm based on SVD which is accurate as well as highly stable to shilling, and combines it with SVD based-CF is described, which offers significant improvement over previous Robust Collaborative Filtering frameworks.
A Survey of Attack-Resistant Collaborative Filtering Algorithms
TLDR
Previous approaches to robust collaborative filtering are reviewed and promising recent approaches that exploit a Singular Value Decomposition (SVD) and are both accurate as well as highly stable to shilling are described.
Collaborative Filtering with Noisy Ratings
TLDR
In NORMA, an adaptive weighting strategy is proposed to decrease the gradient updates of noisy ratings, so that the learned MA models will be less prone to the noisy ratings.
Robust Collaborative Recommendation by Least Trimmed Squares Matrix Factorization
  • Zunping Cheng, N. Hurley
  • Computer Science
    2010 22nd IEEE International Conference on Tools with Artificial Intelligence
  • 2010
TLDR
A least trimmed squares basedMF (LTSMF) is proposed to help improve the robustness of the least squares based MF (LSMF) models, which are shown to be more robust than least squares and another popular robust method - M-estimator.
A Robust Bayesian Probabilistic Matrix Factorization Model for Collaborative Filtering Recommender Systems Based on User Anomaly Rating Behavior Detection
TLDR
A robust Bayesian probabilistic matrix factorization model is constructed for collaborative filtering recommender systems by incorporating the detection of user anomaly rating behaviors and based on which a robust collaborative recommendation algorithm is devised.
Robust recommendation algorithm using an iterative group-based reputation
TLDR
A robust collaborative recommendation algorithm that is based on users' reputations, which is determined based on their grouping behaviours, which has high recommendation accuracy and stronger robustness and does not destroy the user distribution of the original dataset.
Recommending under Threat of Manipulation 7 April
  • Computer Science
  • 2013
TLDR
Simulation results suggest that alternating recommendation between trimmed and non-trimmed methods serves to decrease the effectiveness of manipulation attacks while minimizing the impact on recommendation quality.
A Unified Framework for Reputation Estimation in Online Rating Systems
TLDR
A unified framework for computing the reputation score of a user, given only users' ratings on items is proposed and it is shown that previously proposed reputation estimation methods can be captured as special cases of this framework.
Analysis of robustness in trust-based recommender systems
TLDR
It is argued that exposing ratings presents a robustness vulnerability in trust-based recommender systems and it is shown how this vulnerability can be exploited in the development of profile injection attacks.
...
...

References

SHOWING 1-10 OF 27 REFERENCES
Model-Based Collaborative Filtering as a Defense against Profile Injection Attacks
TLDR
This paper considers two recommendation algorithms, one based on k-means clustering and the other based on Probabilistic Latent Semantic Analysis (PLSA), and shows that, particularly, the PLSA-based approach can achieve comparable recommendation accuracy.
Detecting noise in recommender system databases
TLDR
A framework that enables the detection of noise in recommender system databases and devise techniques that enable system administrators to identify and remove from the recommendation process any such noise that is present in the data.
Collaborative recommendation: A robustness analysis
TLDR
This work analyzes the robustness of collaborative recommendation: the ability to make recommendations despite (possibly intentional) noisy product ratings, and formalizes recommendation accuracy in machine learning terms and develops theoretically justified models of accuracy.
Preventing shilling attacks in online recommender systems
TLDR
Several metrics for analyzing rating patterns of malicious users are proposed and an algorithm for protecting recommender systems against shilling attacks is evaluated that can be employed for monitoring user ratings and removing shilling attacker profiles from the process of computing recommendations, thus maintaining the high quality of the recommendations.
Lies and propaganda: detecting spam users in collaborative filtering
TLDR
This work provides a simple unsupervised algorithm, which exploits statistical properties of effective spam profiles to provide a highly accurate and fast algorithm for detecting spam.
Shilling recommender systems for fun and profit
TLDR
Four open questions are explored that may affect the effectiveness of shilling attacks on recommender systems: which recommender algorithm is being used, whether the application is producing recommendations or predictions, how detectable the attacks are by the operator of the system, and what the properties are of the items being attacked.
Application of Dimensionality Reduction in Recommender System - A Case Study
TLDR
This paper presents two different experiments where one technology called Singular Value Decomposition (SVD) is explored to reduce the dimensionality of recommender system databases and suggests that SVD has the potential to meet many of the challenges ofRecommender systems, under certain conditions.
An Empirical Analysis of Design Choices in Neighborhood-Based Collaborative Filtering Algorithms
TLDR
An analysis framework is applied that divides the neighborhood-based prediction approach into three components and then examines variants of the key parameters in each component, and identifies the three components identified are similarity computation, neighbor selection, and rating combination.
Analysis and Detection of Segment-Focused Attacks Against Collaborative Recommendation
TLDR
This chapter explores an attack model that focuses on a subset of users with similar tastes and shows that such an attack can be highly successful against both user-based and item-based collaborative filtering.
Latent semantic models for collaborative filtering
TLDR
A new family of model-based algorithms designed for collaborative filtering rely on a statistical modelling technique that introduces latent class variables in a mixture model setting to discover user communities and prototypical interest profiles.
...
...