Business Context
Bitcoin OTC is an online platform where users can trade Bitcoin (BTC) or pay for services using BTC
All transactions are peer-to-peer
Trust established via rating scale (-10 to 10)
Important for users to identify other trusted users
Risk: Scammers could create shill accounts to self-rate
Questions/Source of Data
Questions:
Which users are colluding to increase ratings to scam legitimate BTC buyers?
Which users have a legitimate and trustworthy reputation?
Source of Data – Bitcoin OTC Web of Trust:
https://bitcoin-otc.com/viewratings.php
About 6K users rated in database by others
About 32K positive ratings and 3,400 negative ratings
Data Description
Web of Trust Transactional Ratings Data Example:
Interpreting data above:
Hobodave has limited positive ratings (1-2) with small transactions per notes
Wspnut and hobodave gave reciprical +10 (best) ratings. Possible shill rating.
Wspnut has apparently later defrauded rg and received a -10 (worst) rating.
Node Attributes/Links
Selected users with both extremes +10 and -10 scores in history
Directional nodes (Rater > Rated)
Nodes Legend:
Blue Dot : Rater has positive total score
Orange Dot: Rater has negative total score
Black Dot : Rater has neutral (0) total score
Edges Legend:
Green Line : Rating >5 (Great)
Blue Line : Rating 1-5 (Good)
Black Line : Rating 0 (Neutral)
Yellow Line : -1 to -5 (Poor)
Orange Line: -5 to -10 (Very Poor)
A
B
C
Exhibit A: Probably trustworthy – Mostly good ratings from good raters
Exhibit B: Looks like a potential shill fraud ring
Exhibit C: Use caution – a few transactions but suspicious -10 activity
Network Analyses
Shill Network
Good Raters Give -10
Exhibit B
Exhibit A
Exhibit C
Good raters giving very poor scores
Good raters (blue nodes) often giving moderately good feedback (blue or green edges 11 times). Only one -10 and one +10 interaction. More likely to be trustworthy than B or C…
Likely untrustworthy network at the center. Bad raters give overwhelmingly positive feedback, good raters give -10. Some in the center have only one +10 and one -10 (net 0)
Use caution as only good ratings for center node are from bad raters. Good raters give very poor reviews to the center. Also reciprocity of good feedback from bad raters.
Key Findings/Implications for Stakeholders
Findings:
SNA can be used to investigate users in a trust rating database and identify networks of potential shill collusion
Also users could differentiate when a user is legitimate showing mostly positive ratings from good trusted raters
Implications:
Without SNA, users could inadvertently trust scammers
Users can identify legitimate, trusted users
How about accounts associated with scammers – there are ten people are setting setup but are not trustworthy.
