Shilling attacks against recommender systems: a comprehensive survey
Citations
128 citations
88 citations
Cites background from "Shilling attacks against recommende..."
...For immediate gain, some malicious users even exagerate their partners’ services while badmouthing their competiors’ services, which are also known as push attack and nuke atack in shilling attacks [13] ....
[...]
78 citations
78 citations
Cites background from "Shilling attacks against recommende..."
...The natural noise is produced by different sources: while the malicious noise is usually associated to user profiles that match certain patterns [34], the natural noise identification is more difficult because it tends to appear in several ways dissimilar to each other [35]....
[...]
...(1) Malicious noise, associated to noise intentionally introduced by an external agent to bias recommender results [34], and (2) Natural noise, involuntarily introduced by users, and that could also affect the recommendation result [35]....
[...]
75 citations
Cites background from "Shilling attacks against recommende..."
...Indeed, unreliable users have been found in many QoS prediction systems [22]....
[...]
References
5,686 citations
"Shilling attacks against recommende..." refers background or methods in this paper
...(2006b), Hurley et al. (2007), O’Mahony (2004), where shilling attacks are investigated with respect to gains versus attack costs; Resnick and Sami (2008b) examine robust CF approaches....
[...]
...CF system estimates similarities between a and each user in the database, forms a neighborhood by selecting the best similar users, and estimate a prediction (paq) or a recommendation list (top-N recommendation) using a CF algorithm (Herlocker et al. 2004)....
[...]
4,883 citations
4,557 citations
"Shilling attacks against recommende..." refers background or methods in this paper
...Model-based algorithms, on the other hand, first create a model off-line from user-item matrix; they then used that model to produce predictions online (Breese et al. 1998)....
[...]
...Memory-based ones operate over the entire user-item matrix to estimate predictions (Breese et al. 1998)....
[...]
4,299 citations
"Shilling attacks against recommende..." refers methods in this paper
...The term “collaborative filtering” was first coined by the designers of Tapestry (Goldberg et al. 1992), a mail filtering software developed in the early nineties for the intranet at the Xerox Palo Alto Research Center....
[...]
2,401 citations