Graph-Based Anomaly Detection in Decentralized Identity Management Systems Using Temporal Tracking

International Journal of Computer Science and Technology | Vol. 5, No. 4, April 2024 | pp. 1–8

DOI: 10.46882/2024/IJCST/000232

Article Type: Original Research Paper

Title: Graph-Based Anomaly Detection in Decentralized Identity Management Systems Using Temporal Tracking

Names of Authors: Elena Rostova¹, Yuki Tanaka²

Authors’ Affiliations: ¹Data Engineering Laboratory, NordTech University, Oslo, Norway; ²Advanced VLSI Systems Laboratory, Tokyo Institute of Technology, Tokyo, Japan

Abstract: Abstract: Decentralized identity frameworks allow web users to control their personal credentials without relying on centralized single sign-on systems. However, malicious users can exploit these networks by creating multiple fake accounts to skew voting metrics or bypass security limits. This paper presents an automated anomaly isolation framework that maps decentralized identity validation logs using a Temporal Graph Neural Network (TGNN). The model builds real-time structural graphs tracking interaction speeds, link patterns, and credential verification histories. A dynamic edge-weighting algorithm identifies abnormal account groups and separates suspicious nodes based on localized graph changes. We validated this network defense layer using a verification dataset of 12,000 decentralized verification logs under active stress. The system isolated suspicious identity patterns with a 96.4% classification rate. The detection pipeline executed its analysis within a strict timeframe of 32 ms per verification request, proving its viability for live application tracking.

Keywords: Decentralized Identity, Graph Neural Networks, Anomaly Detection, Sybil Attacks, Network Security, Token Verification

Manuscript Timeline: Received: September 12, 2023; Revised: November 20, 2023; Accepted: January 18, 2024; Published: April 12, 2024

Citation: Rostova, E., & Tanaka, Y. (2024). Graph-Based Anomaly Detection in Decentralized Identity Management Systems Using Temporal Tracking. International Journal of Computer Science and Technology, 5(5), 1–8. DOI: 10.46882/2024/IJCST/000232

Subscribe to read the full article: https://internationalscholarsjournals.org/subscribe-to-read