An Auction Reputation System Based on Anomaly Detection

1
An Auction Reputation System Based on Anomaly
Detection
ACM CCS05 (Conf. on Computer and Communications
Security)

• Shai Rubin, Mihai Christodorescu, Vinod
  Ganapathy,
• Jonathon T. Giffin, Louis Kruger and Hao Wang

Computer Sciences Department Univ. of Wisconsin,
Madison
2
Outline

• Motivation find a bargain and avoid a fraud in
  an online auction
• Contribution Identify price inflating sellers
  using three statistical models with anomaly
  detection techniques
• The N model
• The M model
• The P model

3
Data Analyze from eBay

• Collecting data in three weeks period
• Pseudonymous sellers and bidders participate
• Auctions end after a predefined time (e.g. 7
  days)
• Highest bid wins
• Seller sets minimum starting bid
• Shilling a group of bidders that place fake bids
  to inflate the final price

4
Step 1 Average Number of Bids

• What is an indication that prices are high?
• High number of bids
• Goal
• Identify sellers with abnormally high number of
  bids

• 95 of high-volume sellers have less than 7 bids
  per auction
• Model is insensitive to the number of auctions
  posted by a seller

5
Step 1 The N Model

• Correlation
• Many auctions implies low number of bids
• Suspicious seller
• One that posts many auctions and still attracts
  many bids

6
Outline

• Motivation find a bargain and avoid a fraud in
  an online auction
• Contribution anomaly detection system to
  identify price inflating sellers
• The N model a seller is suspicious if he posts
  many auctions that attract many bids
• The M model
• The P model

7
Step 2 Average Minimum Starting Bid

• Legitimate explanation for high number of bids
• Low minimum starting bid
• Goal
• Identify sellers with abnormally high number of
  bids and high minimum starting bid
• Question
• How do we know that the minimum bid is high?

If RMB closes to zero, minimum starting bid is
high
8
Step 2 The M Model

• Correlation
• Low minimum starting bid implies many bids
• M suspicious seller
• Starts with high minimum bid and attracts many
  bids
• MN suspicious seller
• Starts with high minimum bid, posts many
  auctions, and still attracts many bids

9
Outline

• Motivation find a bargain and avoid a fraud in
  an online auction
• Contribution anomaly detection system to
  identify price inflating sellers
• The N model a seller is suspicious if he posts
  many auctions that attract many bids
• The M model a seller is suspicious if he starts
  with high minimum bid and attracts many bids
• The P model

10
Step 3 Bidders Profile of a Seller

• Fraudulent explanation for high number of bids
• Shilling
• Goal
• Identify group of bidders that repeatedly bid and
  lose in a sellers auctions
• Suspicious seller
• N sellers with abnormally high number of bids
  and
• M high starting bid and
• P has a group of bidders that repeatedly bid and
  lose

11
Bidder presence/win curves

• The abnormal seller

• 5 of bidders participate in 95 of the auctions

• The same 5 of bidders participate in 95
  auctions only won 10 of bids

12
Step 3 The P Model (2/2)

• The normal seller

13
Outline

• Motivation find a bargain and avoid a fraud in
  an online auction
• Contribution anomaly detection system to
  identify price inflating sellers
• The N model a seller is suspicious if he posts
  many auctions that attract many bids
• The M model a seller is suspicious if he starts
  with high minimum bid and attracts many bids
• The P model a seller is suspicious if he has a
  group of bidders that repeatedly participate and
  lose

14
Threat Analysis (1/2)

• Normalizing a N score
• Using fewer shill bids per auction
• Creating a new identity
• Adding shill bids to auctions of other sellers
• Normalizing a M score
• Decreasing average number of bids discussed in
  normalizing a N score
• Decreasing minimum starting bid

15
Threat Analysis (2/2)

• Normalizing a P score
• Distributing shill bids
• Letting some shill bidders win

16
(No Transcript)
17
Conclusion

• Address a concern that ignored by current
  reputation system
• Demonstrate the efficacy and practicality of
  these ideas by building a reputation system and
  using it on real-world data gathered from eBay