Brent Smith’s scientific contributions

... In today's era of information explosion, recommender systems (RecSys) effectively assist users in filtering out uninteresting information and providing tailored services, which are widely applied in various scenarios such as e-commerce [26,36,15], streaming platforms [30,13,4], and social media [10, 6, 5]. For instance, Amazon's recommender system utilizes user's historical purchase records, browsing behaviors, and data from other users to personalize product recommendations, helping users discover items they may be interested in but have not yet found and enhancing user experience [16,21]. Recently, Large Language Models (LLMs) have fundamentally revolutionized existing recommender systems due to their powerful language comprehension capabilities and rich open-world knowledge [7, 38,43]. ...

Reference:

Exploring Backdoor Attack and Defense for LLM-empowered Recommendations

... These attacks have been developed for a range of systems, including graph-based [24], association rule-based [25], matrix factorization-based [26,27], and neighborhood-based [28] recommender systems. For example, O'Mahony et al. [18] studied the robustness of user-based collaborative filtering (CF) methods [29] by injecting fake users. Burke et al. [30] analyzed the impact of low-knowledge attack strategies on CF methods, aiming to promote or demote items. ...

Reference:

DIFshilling: A Diffusion Model for Shilling Attacks