Fighting fire with fire

Adapting to the evolution of fraud techniques

The COVID-19 pandemic led to an unprecedented surge in US unemployment. Coupled with broadened eligibility and enhanced funding for unemployment insurance (UI), this surge attracted a new breed of more sophisticated fraudsters. Individual bad actors and organized crime rings are aggressively targeting UI agencies with large-scale, technology-enabled fraud schemes. These agencies are not alone: The threat extends to all benefits systems that have high-volume funding distribution, across the country and at every government level.

The surge in claims presented state agencies with a wide range of challenges. They were in uncharted territory with new eligibility requirements, non-traditional applicants, and shifting guidance on program rules. At the same time, they faced critical staffing and workload issues.

Public benefits agencies should consider acting now to deploy and rapidly scale AI and machine learning that keeps up with the evolution of fraud techniques. Fraudsters often leave behind digital fingerprints in the form of anomalous patterns and behaviors. An agency can use anomaly detection, machine learning, and network analytics to understand normal behavior as well as frequency and order of account activities to identify potentially fraudulent activity. With these and other tools, agencies can look beyond known tactics so they can keep up with fraudsters in adapting to new and emerging schemes.

Techniques include:

• Statistical methods to identify unique activities that are anomalous to specific user groups—that is, activities that do not align with a user’s typical interaction with a claim or claim attributes. Examples include home addresses, bank account information, etc.
• Machine learning techniques to label users who share patterns that resemble known insider threats.
• Anomaly detection to analyze suspicious internal activities—for example, actions taken on dormant claims and the frequency of these actions.
• Network analytics to determine the connectivity between internal users and claimants. Network graphs can help identify common attributes and hidden connections between claimants and users.

By fighting back with AI tools and techniques, UI agencies can reduce the amount of data exfiltrated and fraud committed. However, even with the most sophisticated detection approaches, UI agencies cannot eliminate fraud entirely. Even as unemployment recedes, claims volume will likely surge again in response to future recessions, economic shocks, and the continued evolution of fraud techniques. By being proactive with advanced fraud detection measures, agencies can make progress in preventing the proliferation of these crimes and help ensure that eligible candidates receive vital public benefits.

About Deloitte AI Institute for Government
The Deloitte AI Institute for Government is a hub of innovative perspectives, groundbreaking research, and immersive experiences focused on artificial intelligence (AI) and its related technologies for the government audience. Through publications, events, and workshops, our goal is to help government use AI ethically to deliver better services, improve operations, and facilitate economic growth.

Get in touch

	Scott Malm Principal Deloitte Consulting LLP +1 651 246 5075
	Carol Tannous Managing Director Deloitte Transactions and Business Analytics +1 617 437 3872