Robustheit in Empfehlungssystemen
•
0 likes•777 views
The document discusses various types of attacks on recommender systems, such as profile injection attacks where malicious users try to manipulate the recommendations. It analyzes the robustness of collaborative filtering recommender algorithms to different attack strategies like random attacks, bandwagon attacks, and segment attacks. The results show that collaborative filtering is vulnerable to profile injection attacks but that its performance can be improved through detection and defense techniques that identify suspicious profiles and reduce their influence on the recommendations.
Robustheit in Empfehlungssystemen
- 5. 5
- 6. ï‚¢ ï‚¢ 6
- 7. ï‚¢ ï‚¢ 7
- 8. ï‚¢ ï‚¢ ï‚¢ 8 Burke, Mahony, Hurley, Robust Collaborative Recommendation
- 11.    ∅  11 Williams, Mobasher, Burke, Defending Recommender Systems: Detection of Profile Injection Attacks, 2007
- 22. ï‚¢ ï‚¢ ï‚¢ 22
- 25. 1,80 1,60 1,40 1,20 1,00 0,80 0,60 0,40 0,20 0,00 Random Attack Bandwagon Attack Average Attack 25 Burke, Mahony, Hurley, Robust Collaborative Recommendation
- 26. 0,90 0,80 0,70 0,60 0,50 0,40 0,30 0,20 0,10 0,00 All-Users Segment Attack 26 Burke, Mahony, Hurley, Robust Collaborative Recommendation
- 27. 0,00 -0,50 Reverse Bandwagon Average Attack Random Attack Bandwagon Love/Hate Attack Attack -1,00 -1,50 -2,00 -2,50 27 Burke, Mahony, Hurley, Robust Collaborative Recommendation
- 28. 0,10 0,00 -0,10 -0,20 -0,30 -0,40 -0,50 -0,60 -0,70 -0,80 Reverse Bandwagon Average Attack Random Attack Bandwagon Love/Hate Attack Attack 28 Burke, Mahony, Hurley, Robust Collaborative Recommendation
- 29. 80% 70% 60% 50% 40% 30% 20% 10% 0% Average Attack Probe Attack Popular Attack 29 Burke, Mahony, Hurley, Robust Collaborative Recommendation
- 34. ï‚¢ ï‚¢ 34 Williams, Mobasher, Burke, Defending Recommender Systems: Detection of Profile Injection Attacks, 2007
- 35. ï‚¢ ï‚¢ ï‚¢ 35
- 36. ï‚¢ ï‚¢ ï‚¢ 36
- 43. ï‚¢ ï‚¢ ï‚¢ 43
- 44. 44