Identifying Multi-ID Users in Open Forums

1. Identifying Multi-ID Users in Open Forums Hung-Ching Chen Mark Goldberg Malik Magdon-Ismail

2. Who is Using Multiple IDs? Consider two chatrooms, “letters” and “numbers” letters numbers dogg: hey anna, Z rocks! anna: hey dogg whats new mack: what’s that dogg? dogg: not much dogg: Z is the best mack: i love 5 catt: i don’t joop: 27 rules! catt: i agree joop :)

3. Multi-ID Users letters numbers dogg: hey anna, Z rocks! anna mack dogg/catt joop

4. Multi-ID Users letters numbers dogg: hey anna, Z rocks! mack: i love 5 anna mack dogg/catt joop

5. Multi-ID Users letters numbers dogg: hey anna, Z rocks! anna: hey dogg whats new mack: i love 5 anna mack dogg/catt joop

6. Multi-ID Users letters numbers dogg: hey anna, Z rocks! anna: hey dogg whats new mack: what’s that dogg? mack: i love 5 anna mack dogg/catt joop

7. Multi-ID Users letters numbers dogg: hey anna, Z rocks! anna: hey dogg whats new mack: what’s that dogg? mack: i love 5 catt: i don’t anna mack dogg/catt joop

8. Multi-ID Users letters numbers dogg: hey anna, Z rocks! anna: hey dogg whats new mack: what’s that dogg? dogg: not much mack: i love 5 catt: i don’t anna mack dogg/catt joop

9. Multi-ID Users letters numbers dogg: hey anna, Z rocks! anna: hey dogg whats new mack: what’s that dogg? dogg: not much mack: i love 5 catt: i don’t joop: 27 rules! anna mack dogg/catt joop

10. Multi-ID Users letters numbers dogg: hey anna, Z rocks! anna: hey dogg whats new mack: what’s that dogg? dogg: not much dogg: Z is the best mack: i love 5 catt: i don’t joop: 27 rules! anna mack dogg/catt joop

11. Multi-ID Users letters numbers dogg: hey anna, Z rocks! anna: hey dogg whats new mack: what’s that dogg? dogg: not much dogg: Z is the best mack: i love 5 catt: i don’t joop: 27 rules! catt: i agree joop :) anna mack dogg/catt joop

12. Overview • A model of a public forum • Two efficient statistics-based algorithms for identification of multi-ID users • Simulation results

13. Overview • A model of a public forum • Two efficient statistics-based algorithms for identification of multi-ID users • Simulation results

14. A Model of a Public Forum • Every actor has one response queue • Common average and variance of response delay • The server has a queue to process messages with very short delay

15. Model Friendship Graph anna mack dogg joop catt

16. Model Alias Graph anna mack dogg joop catt

17. Multi-ID Users • One response queue for each actor dogg catt anna catt anna mack dogg/catt joop

18. Multi-ID Users letters numbers dogg: hey anna, Z rocks! dogg dogg catt anna catt anna mack dogg/catt joop

19. Multi-ID Users letters numbers dogg: hey anna, Z rocks! mack: i love 5 mack dogg dogg catt catt anna mack dogg/catt joop

20. Multi-ID Users letters numbers dogg: hey anna, Z rocks! anna: hey dogg whats new mack: i love 5 anna dogg dogg mack catt anna mack dogg/catt joop

21. Multi-ID Users letters numbers dogg: hey anna, Z rocks! anna: hey dogg whats new mack: what’s that dogg? mack: i love 5 mack anna dogg mack catt anna mack dogg/catt joop

22. mack anna Multi-ID Users letters numbers dogg: hey anna, Z rocks! anna: hey dogg whats new mack: what’s that dogg? mack: i love 5 catt: i don’t catt mack catt anna mack dogg/catt joop

23. Multi-ID Users letters numbers dogg: hey anna, Z rocks! anna: hey dogg whats new mack: what’s that dogg? dogg: not much mack: i love 5 catt: i don’t dogg mack catt anna catt anna mack dogg/catt joop

24. Multi-ID Users letters numbers dogg: hey anna, Z rocks! anna: hey dogg whats new mack: what’s that dogg? dogg: not much mack: i love 5 catt: i don’t joop: 27 rules! joop dogg catt mack catt anna mack dogg/catt joop

25. Multi-ID Users letters numbers dogg: hey anna, Z rocks! anna: hey dogg whats new mack: what’s that dogg? dogg: not much dogg: Z is the best mack: i love 5 catt: i don’t joop: 27 rules! dogg joop dogg catt mack anna mack dogg/catt joop

26. Multi-ID Users letters numbers dogg: hey anna, Z rocks! anna: hey dogg whats new mack: what’s that dogg? dogg: not much dogg: Z is the best mack: i love 5 catt: i don’t joop: 27 rules! catt: i agree joop :) catt dogg dogg catt joop anna mack dogg/catt joop

27. Overview • A model of a public forum • Two efficient statistics-based algorithms for identification of multi-ID users • Simulation results

28. First Algorithm • Collect information {‹timei, IDi›} • Compute minimum time delay minD for every pair of IDs • Cluster the delays using k-means into two groups. • Call pairs with larger center suspected to be the same actor (red) • Call pairs with smaller center suspected to different (blue) • Connected components using red edges are ID groups representing one actor

29. 1. Collect information time dogg dogg dogg mack mack anna catt catt joop

30. 2. minD(dogg, mack) dogg dogg dogg mack mack

31. 2. minD(dogg, catt) dogg dogg dogg catt catt

32. 3. Cluster minD values minD(dogg, catt) minD(dogg, mack) Different actors Same actor

33. Contradictions 3. Resulting Alias Graph anna mack dogg joop catt

34. 4. One Red Component anna mack dogg joop 9 false positives 10% accuracy catt

35. Refinement Algorithm • Find groups connected with red edges • Color IDs into smaller ID groups using blue edges Follow original algorithm

36. 4. One Red Component anna mack dogg joop catt

37. 5. Color Blue Graph anna mack dogg joop catt

38. 1 false positive 90% accuracy 5. Color Blue Graph anna mack dogg joop catt

39. Overview • A model of a public forum • Two efficient statistics-based algorithms for identification of multi-ID users • Simulation results

40. Simulation results

41. Future Work • Real-life testing • More sophisticated model • Different types of users • Short and overlapping online appearances • Length of posts • Algorithm improvements • Statistical evaluation of the new models • Lexical and semantic analysis