- Necmettin Erbakan Üniversitesi Fen ve Mühendislik Bilimleri Dergisi
- Cilt: 5 Sayı: 2
- Shilling Attack Detection with One Class Support Vector Machines
Shilling Attack Detection with One Class Support Vector Machines
Authors : Halil Ibrahim Ayaz, Zehra Kamişli Öztürk
Pages : 246-256
Doi:10.47112/neufmbd.2023.22
View : 56 | Download : 84
Publication Date : 2023-12-31
Article Type : Research
Abstract :Recommender systems play a vital role in various online platforms, assisting users in discovering new products, services, and content considering their preferences. However, these systems are vulnerable to manipulation through shilling attacks, where malicious users artificially inflate or deflate ratings, leading to biased recommendations. It is crucial to emphasize the importance of researching, understanding, and mitigating these attacks. Detecting such attacks is crucial to maintaining the integrity and effectiveness of recommender systems. In the literature, lots of studies are presented to detect shilling attacks. The most well-known clustering methods are adapted for different attack models. This paper explores using One-Class Support Vector Machines (OCSVM) as a robust technique for detecting shilling attacks. One-Class SVMs are a specialized variant of the traditional Support Vector Machines, primarily designed for anomaly and novelty detection tasks. MovieLens100K dataset is used to validate the proposed method. As a result, precision and recall values are given for different attack and filler sizes.Keywords : Öneri sistemleri, Şilin atakları, Tek sınıflı destek vektör makinaları