1 Introduction
2 Background and Objectives
2.1 Online Trust and Reputation Systems
2.2 Exploiting Reputation Systems
Attack class | Name | Description |
---|---|---|
Unfair feedback | Ballot stuffing | The attacker provides many high ratings to unfairly push the reputation of an entity |
Bad mouthing | The attacker provides many low ratings to unfairly destroy the reputation of an entity | |
Inconsistent behavior | Value imbalance exploitation | The attackers gathers good reputation selling cheap items, but at the same time cheats on the expensive ones |
On-off attack | The attackers first acts honestly to build high reputation, than “milks” the good reputation. After a certain threshold value is reached, he behaves honestly again and starts from the beginning | |
Identity-based | Whitewashing | The attacker behaves maliciously from the beginning. After he has received negative ratings, he opens a new account |
Sybil-attack | The attacker creates many accounts (Sybils) at the same time to increase his influence in a community |
2.3 Robust Metrics vs. Transparent Presentation
Name | Author | Formal basis |
---|---|---|
Beta reputation system, TRAVOS | Beta probability density functions (PDF) | |
iClub |
Liu et al. (2011) | Clustering-based model |
Evidential model |
Yu and Singh (2002) | Belief model, Dempster-Shafer theory |
Web services reputation |
Malik et al. (2009) | Hidden Markov Model |
REGRET |
Sabater and Sierra (2001) | Fuzzy model |
2.4 Research Gap and Objectives
2.5 Research Approach
3 Interactive Reputation Systems
3.1 Process Model
3.2 Conceptual Design
3.2.1 Computation
3.2.2 Presentation
3.2.3 Integrating Computation and Presentation
3.3 Implementation
3.3.1 A Generic Software Framework
3.3.2 Prototype: Extension of an eBay-like Feedback Profile
4 Evaluation
4.1 Case Study
4.2 International User Study
Measure | Old interface | New interface |
---|---|---|
Correct detection of malicious seller | 56% | 77% |
Preferences for honest (of user who decided to buy) | 58% | 85% |
Preferences for malicious (of user who decided to buy) | 30% | 7% |
No preference (of user who decided to buy) | 12% | 8% |
Sensemaking Score [0; 1] | 0.25 | 0.46 |