Enhancing popSAD: A New Approach to Shilling Attack Detection in Collaborative Recommenders

Maintaining the integrity and reliability of user-generated content requires the detection of shilling profiles or attack profiles in recommender systems. By embracing the concepts of Popularity, Quantity User Diversity (QUD), Recency User Diversity (RUD), and Mean Popularity of Items (MUP) as essential attributes, we present a unique technique for shilling profile detection in collaborative filtering recommender systems in this study. Our approach attempts to improve the precision and dependability of the shilling profile identification algorithms. We run tests on the ML-100K dataset to assess how well our strategy performs using accuracy, recall, and F1-score measures. The findings show how well our suggested method performs in terms of accuracy, recall, and F1-score values, underscoring how well it can recognize shilling characteristics. This study advances the subject of shilling profile recognition in recommender systems by presenting a thorough strategy that integrates several variables for increased detection precision.