safe-graph/graph-fraud-detection-papers

What it solves

The project addresses the challenge of keeping up with the fast-growing research on fraud, anomaly, and outlier detection using graph- and transformer-based methods. It organizes a large set of academic papers and related resources into a single, searchable collection so researchers and practitioners can quickly find relevant work without scouring multiple sources.

How it works

The repository is a structured list of papers grouped by year and by category. It has dedicated sections for LLM/Transformer papers and deep-learning graph papers (sorted by year), as well as sections for non-deep-learning graph papers, toolboxes, datasets, surveys, and other resources. Each entry includes the paper's title, venue, a link to the paper, and often a link to its code. In addition, the project provides an interactive dashboard for browsing, filtering, and searching the papers, and a local RAG-based LLM chatbot pre-loaded with over 250 papers to support deep research.

Who it's for

Fraud and anomaly detection researchers, data scientists, machine learning engineers, and students who want to stay current with academic literature and want quick access to papers, code, and datasets in this domain. It is also useful for anyone building systems that need to detect fraudulent or unusual behavior in graph-structured data.

Highlights

• Curated lists of graph- and transformer-based papers for fraud, anomaly, and outlier detection.
• Sections covering LLM/Transformer methods, deep-learning graph methods, non-deep-learning methods, toolboxes, datasets, and surveys.
• Each paper entry includes links to the paper and often to its code.
• An interactive dashboard to view, filter, and search the papers.
• A local RAG-based LLM chatbot ("paper_chatbot") that has 250 publicly accessible papers for deep research.