Synthetic identity fraud is an uprising concerning risk to multiple industries and the loss for unsecured credit products is estimated to hit $2.4 billion in 2023. One of the most challenges in synthetic identity fraud detection is due to the nature of the problem: fraudsters are able to create fake identities with consistent identity information, and many times with even a matched social media profile and employment history.
In this section, I would like to reveal a unique data science framework to detection the synthetic identities by first building a scalable digital identity graph network with close to a billion of edges, attribute the suspicious nodes with fraudulent signals derived with novelty detection algorithms, and then apply the graph mining algorithms to detect suspicious communities and finally recommend the top risk nodes and subgraph for synthetic identity fraud detection.
The section is not technical driven, which means that even if you are not a graph computing engineer you will be able to follow through the machine learning approach and framework in the presentation. This section combines the high level use case illustration as well as the detailed methodology articulation to explain how to set up the graph network and apply the graph mining algorithms for suspicious community detection.