PulseAugur
EN
LIVE 07:34:17

Researchers develop Sharpness-Aware Poisoning to improve attack transferability in recommender systems.

By PulseAugur Editorial · [2 sources] ·

Researchers have developed a new attack method called Sharpness-Aware Poisoning (SharpAP) to improve the transferability of malicious data injections in recommender systems. This technique aims to overcome the limitations of current methods that struggle when the surrogate model used for attack preparation differs structurally from the actual target model. SharpAP seeks an approximate worst-case victim model to optimize poisoned data, making it more robust and less sensitive to variations in model architecture. Experiments on real-world datasets indicate that SharpAP significantly enhances the effectiveness of these attacks. AI

IMPACT Enhances the robustness of data poisoning attacks against recommender systems, potentially impacting platform security and user trust.

RANK_REASON This is a research paper detailing a novel attack method for recommender systems.

Read on arXiv cs.LG →

AI-generated summary · Google Gemini · from 2 sources. How we write summaries →

Researchers develop Sharpness-Aware Poisoning to improve attack transferability in recommender systems.

COVERAGE [2]

  1. arXiv cs.LG TIER_1 English(EN) · Junsong Xie, Yonghui Yang, Pengyang Shao, Le Wu ·

    Sharpness-Aware Poisoning: Enhancing Transferability of Injective Attacks on Recommender Systems

    arXiv:2604.22170v1 Announce Type: new Abstract: Recommender Systems~(RS) have been shown to be vulnerable to injective attacks, where attackers inject limited fake user profiles to promote the exposure of target items to real users for unethical gains (e.g., economic or political…

  2. arXiv cs.LG TIER_1 English(EN) · Le Wu ·

    Sharpness-Aware Poisoning: Enhancing Transferability of Injective Attacks on Recommender Systems

    Recommender Systems~(RS) have been shown to be vulnerable to injective attacks, where attackers inject limited fake user profiles to promote the exposure of target items to real users for unethical gains (e.g., economic or political advantages). Since attackers typically lack kno…

RELATED ENTITIES

RELATED TOPICS