Research map / Adversarial robustness and security / Attacks on text and graphs
Data Poisoning Attacks on Factorization-Based Collaborative Filtering
Read the paperCheck your idea against 108,967 papers
AI-extracted summary. Written by AI from the paper's title only (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
Measured effects reported in the paper
| What was changed | Where | Measured | Effect |
|---|---|---|---|
| Optimize malicious profiles with projected gradient ascent for availability attacks | Alternating minimization · MovieLens 20M | RMSE | Helped |
| Use SGLD to sample malicious profiles from a normal-user-informed posterior instead of PGA | Alternating minimization · MovieLens 20M | RMSE | Hurt |
| Optimize malicious profiles with PGA or SGLD rather than uniform random profiles | Alternating minimization · MovieLens 20M | Attacker utility | Helped |
| Use PGA with uniformly selected rated items rather than normal-user-like item selection | Alternating minimization · MovieLens 20M | Paired t-test p-value: p<0.05 | Hurt |
| Use SGLD to make malicious profiles resemble normal users | Alternating minimization · MovieLens 20M | Paired t-test p-value: larger than 0.7 | Helped |
| Combine availability and integrity goals rather than targeting availability alone | Alternating minimization attack · MovieLens 20M | RMSE | Hurt |
| Target a specific item to manipulate its rating | PGA and SGLD attacks · MovieLens 20M | Average item rating | Helped |
| Push a selected item without an availability-error objective | Alternating minimization integrity attack · MovieLens 20M, selected item | Average item rating: 0.8 to around -0.3 | Helped |
| Reduce selected-item popularity while also reducing system-wide availability error | Alternating minimization hybrid attack · MovieLens 20M, selected item | Average item rating: 0.8 to about -0.1 | Mixed |
| Increase β to prioritize attack utility over profile similarity | SGLD attack · MovieLens 20M | RMSE | Helped |
| Increase β to prioritize attack utility over profile similarity | SGLD attack · MovieLens 20M | Paired t-test p-value | Hurt |
| Combine availability and integrity objectives for selected-item manipulation | Nuclear norm minimization attack · MovieLens subset, selected item | Average item rating | Mixed |
Closest papers by meaning
- Understanding and Improving Adversarial Collaborative Filtering for Robust Recommendation (NeurIPS 2024 · similarity 0.87)
- Revisiting Injective Attacks on Recommender Systems (NeurIPS 2022 · similarity 0.87)
- Revisiting Item Promotion in GNN-Based Collaborative Filtering: A Masked Targeted Topological Attack Perspective (AAAI 2023 · similarity 0.86)
- Coordinated Attacks against Contextual Bandits: Fundamental Limits and Defense Mechanisms (ICML 2022 · similarity 0.84)
- Untargeted Attack against Federated Recommendation Systems via Poisonous Item Embeddings and the Defense (AAAI 2023 · similarity 0.84)
- Sim4Rec: Data-Free Model Extraction Attack on Sequential Recommendation (AAAI 2025 · similarity 0.82)
Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.