Scholarlink

In the realm of collaborative filtering recommender systems, detecting shilling attacks, where malicious users artificially influence recommendations, is a critical challenge. Traditional methods often struggle with the complexity and high-dimensionality of feature spaces. To overcome these limitations, we harness the capabilities of CNN-OPTICS, a novel approach that integrates Convolutional Neural Networks (CNNs) with the OPTICS (Ordering Points To Identify the Clustering Structure) algorithm. CNNs are employed to extract deep, nuanced features from user interaction data, revealing subtle patterns and anomalies, while OPTICS clusters these extracted features, identifying meaningful clusters and detecting anomalies indicative of shilling attacks. This synergy between deep feature extraction and flexible clustering allows CNN-OPTICS to outperform traditional methods, providing better resistance to noise and enhanced detection of malicious behavior. Extensive evaluations on the MovieLens and Netflix datasets confirm that CNN-OPTICS delivers superior performance across various attack types, highlighting the potential of integrating deep learning with clustering techniques to secure such systems.

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Harnessing CNN and OPTICS Algorithm for Detecting Shilling Attacks in Recommender Systems

  • 会议论文

  • 发表日期：

    2025年

查看全文
  • Ouahiba Belgacem,
  • Boudjemaa Boudaa,
  • Abderrahmane Kouadria,
  • Abdelhafid Abouaissa

摘要

In the realm of collaborative filtering recommender systems, detecting shilling attacks, where malicious users artificially influence recommendations, is a critical challenge. Traditional methods often struggle with the complexity and high-dimensionality of feature spaces. To overcome these limitations, we harness the capabilities of CNN-OPTICS, a novel approach that integrates Convolutional Neural Networks (CNNs) with the OPTICS (Ordering Points To Identify the Clustering Structure) algorithm. CNNs are employed to extract deep, nuanced features from user interaction data, revealing subtle patterns and anomalies, while OPTICS clusters these extracted features, identifying meaningful clusters and detecting anomalies indicative of shilling attacks. This synergy between deep feature extraction and flexible clustering allows CNN-OPTICS to outperform traditional methods, providing better resistance to noise and enhanced detection of malicious behavior. Extensive evaluations on the MovieLens and Netflix datasets confirm that CNN-OPTICS delivers superior performance across various attack types, highlighting the potential of integrating deep learning with clustering techniques to secure such systems.