User Tools

Site Tools


10a.006.UL - Application-Domain-Agnostic Anomaly Detection in Blockchain Transaction Graphs

Project - Summary

The popularity of blockchain technology is on the rise due to the boom of Bitcoin and Ethereum. Blockchain technology’s decentralized immutable record keeping has created waves of interest in different application areas. The introduction of smart contracts has expanded the horizon for blockchains. However, misuse of blockchain features threaten the acceptance of blockchain technology. The prevalence of fraud and money laundering in cryptocurrency blockchain applications are the most common examples of this misuse. We cover the deficiencies of current research by developing an agnostic approach that accounts for the evolving and unpredictable patterns misuse presents in the data regardless of the blockchain technology application area. We employ graph theory and robust principal component analysis to take advantage of blockchain graph properties to identify, monitor and visualize anomalous transaction subgraphs.

Project - Team

Team Member Role Email Phone Number Academic Site/IAB
Mehmet Engin Tozal PI (337) 482 6604 UL Lafayette
Somtoo Chuckwu Student TBA UL Lafayette

Project - Novelty of Approach

Anomaly detection in blockchain graphs require new approaches because of their intrinsic features:

  • Large graph size
  • Disallowed node deletions
  • Frequent node/edge additions
  • Shorter anomaly spans
  • Varying anomaly patterns

The proposed approach is novel in terms of both exploiting the subgraph embedding through RPCA and lower memory and computational requirements to support agile anomaly detection and visualization under frequent graph updates.

Project - Deliverables

1 Develop a user-friendly visualization to allow users to explore anomalous subgraphs further
2 Develop the proposed anomaly detection approach and explore others
3 Fine tune the parameters of RPCA and sliding window size to improve the overall performance

Project - Benefits to IAB

Anomaly detection in blockchain has been studied by various researchers. Various approaches have been studied, but they face a few setbacks. The main setbacks are a lack of consensus on what constitutes an anomaly in a blockchain application, a lack of ground truth to confirm the findings of the studies and the size of data involved in blockchain applications. Previous studies also have not provided a generalized anomaly detection model that can be applied across blockchain applications.

In addition to these setbacks, an anomaly detection model must take a blockchain application’s large ledger size, high dynamicity, forbidden node/edge deletion and frequent node/edge additions, evolving anomalous behaviors over shorter periods of time and capricious user behavior patterns into consideration. These blockchain characteristics can be overwhelming to a user/observer trying to study/investigate the blockchain application. Our system combats this by providing a tool for users to ease the strain. Our system can be applied across various blockchain applications to identify anomalous subgraphs that are involved in some sort of misuse. The system represents the results/findings in a visual medium to the user to aid in further investigation.

Project - Documents

projects/year10/10a.006.ul.txt · Last modified: 2022/10/03 15:19 by sally.johnson