Defending Trust: Innovations in First-Party Fraud Prevention
Did you know First-party fraud is a growing $100 billion problem?
Did you know First-party fraud is a growing $100 billion problem? First-party fraud (also known as ‘Friendly Fraud’), where individuals use their own personal information to knowingly and deceitfully commit fraud for financial gain. This is pervasive, with 35% of Americans in surveys admitting to committing some form of first-party fraud, often driven by economic hardship1. The challenge with first-party fraud is that it is much harder to detect than identity theft or synthetic fraud, as the perpetrator is using their own legitimate personal information.
This whitepaper explores the type of First-Party Fraud, current market trends, challenges with existing fraud solutions and case study on how VALID AI-powered solutions are helping fintech, banks to combat fraud and proactively mitigate fraud risks.
First-party fraud poses a significant threat to financial institutions, resulting in substantial losses and reputational damage. Traditional approaches to combat this issue often fall short due to the fragmented nature of data and the lack of comprehensive insights into fraudulent activities. In response, the establishment of a First-Party Fraud Data Consortium emerges as a promising solution.
First-party fraud, a rapidly escalating threat, involves individuals knowingly misusing their own legitimate personal information to deceive businesses and financial institutions for illicit financial gain, often driven by economic hardship. Unlike identity theft, where perpetrators steal someone else's identity, first-party fraud is committed by the individual themselves through misrepresentation of details like income, employment status, or intent to repay loans. Common tactics include application fraud, sleeper fraud, chargeback abuse, and refund fraud. Detecting and preventing first-party fraud poses unique challenges as perpetrators use their real identities, making it harder to distinguish from legitimate consumer behavior. Combating this escalating $100 billion problem requires a multi-layered strategy - leveraging advanced analytics, machine learning models trained on cross-industry insights, robust identity verification, and tailored fraud mitigation policies.
There are common types of first-party fraud include2:
-
Sleeper fraud: Opening a loan account with no intention of ever repaying the debt, only to disappear after making a few payments
- Fronting: Setting up a service or account in another person's name, such as a young driver getting cheaper car insurance by applying under a parent's name
- Address fronting: Using a different, lower-risk address, like a holiday home, to get cheaper services like loan requests, car insurance
- Chargeback fraud (or "cyber shoplifting"): Where a customer makes a purchase but then disputes the charge with their bank or credit card company, claiming they didn't receive the item or it wasn't as described, to get a refund while keeping the product.
- Refund and promotion abuse: Customers abuse return and promotion policies to get products for free or at a heavily discounted price, such as by returning items after using them or claiming multiple discounts on a single purchase.
- Wardrobing (or "de-shopping"): Purchasing items, often clothing, with the intention of using them and then returning them for a full refund.
There are various interesting statistics identity and first party fraud verification.
- First-party fraud costs U.S. financial institutions and merchants more than $100 billion annually
- 35% of Americans admit to committing first-party fraud themselves
- 25% of the total consumer charge-offs in US are from first party fraud3
- 42% of Gen Z consumers are found to be engaging in first-party fraud4
Leveraging data consortiums and pooling data from multiple sources to identify patterns of first-party fraud across different platforms and accounts. This helps detect fraudulent behavior that may not be visible from a single company's data.
Using advanced analytics and machine learning models trained on consortium data to accurately identify first-party fraud risk during account opening and onboarding processes.
Fraudsters are exploiting account funding and/or item verification processes by providing false information or misrepresenting their financial circumstances to obtain services or credit lines. Common tactics include income exaggeration, employment fabrication, and address fronting
Originations: First party fraud involves an individual who makes a promise of future repayment in exchange for goods and services without the intent to repay.
Account Management: Continuously monitoring customer behavior and account activity patterns post-lending to detect potential first-party fraud bust-outs before full credit lines are maxed out.
Individuals open multiple lines of credit or loans across different institutions using their real identities. They then max out the credit limits with no intention of repaying, a practice known as "loan stacking" or "bust out" schemes.
- Difficulty in Detection: First-party fraud is inherently difficult to detect because fraudsters use their own legitimate identities and credentials
- Lack of Historical Data: Many financial services providers often lack the historical data and fraud patterns that traditional banks can analyze to define and detect first-party fraud behaviors
- Lack of Data signals: Detecting first-party fraud requires analyzing alternative data signals and behavioral patterns across multiple platforms and industries, which a single company's data may not reveal. This has led to the formation of data consortiums to pool insights.
- Evolving Fraud Tactics: As solutions are developed, fraudsters continually adapt their tactics, making it an ongoing challenge to stay ahead of evolving first-party fraud schemes
- Manual Processes and Friction: Relying on manual processes and introducing excessive friction during account opening or transactions can lead to customer abandonment and lost revenue
Valid Edge is an AI-powered data consortium, a game-changer solution to combat identities and fraud detection challenges. Valid Edge comprises of 300M+ unique bank accounts with $6T annual check transaction volume in 2023, which is 22% of $27T total check payments in US per Federal Reserve Payments Study (2021)5. By aggregating data from multiple sources, primarily from 7 out of top 25 banks, the consortium fosters a collaborative environment for fraud detection and prevention. Leveraging advanced analytics and AI-powered machine learning techniques, the consortium enables real-time identification of suspicious behaviors and patterns, empowering members to proactively mitigate fraud risks.
- Real-time Account and Transaction Verification: Allows members to query the consortium data in real-time, cross-reference it with their own risk models, and verify the legitimacy of senders, receivers, or counterparties involved in transactions. This capability strengthens identity verification and risk assessment processes
- Improved Fraud Detection through Shared Data and Analytics: By pooling data from multiple companies, a fraud consortium enables members to gain insights into fraud patterns and suspicious activities that may not be visible from any single organization's data alone.
- Cost Savings and Resource Optimization: Allows companies to share resources, collaborate, knowledge share, expertise, and the costs associated with fraud detection, intervention, and recovery efforts
Valid Edge Case Study: Identify Fraudulent Checks
Problem Statement: There is a growing check fraud in banks, increasing 50%+6 YoY. It is critical for banks to identify the fraudulent checks and ways to reduce fraud $ losses
Clients: 3 out of top 20 banks
VALID has developed ML-algorithm based fraud consortium and leveraged behavior analytics to study 3 out of top 20 banks check deposits activity of 2M+ checks worth $3T and predict the likelihood of fraudulent checks.
By performing the analysis based on the activity on check numbers, VALID classified the checks into high vs. low risks level based on the check deposit transactions.
| Category | High Risk | Low risk | |
| % of total items | 22% | 78% | |
| % of total returns ($) | 68% | 32% | |
| Likelihood of return | 5X | 1X |
The check deposit data was used from 3 out of the top 20 banks. This included ~2M checks with $3T+ check deposit amount from the 3 months of transaction activity. These 2M checks were categorized into 6 risk grades based on check number range deviation grades as shown below. The check number deviation grade is defined based on the comparison of check number to the min/max historical check numbers from the payer.
| Risk category | Check Number Pange deviation |
definition |
|
| High Risk |
-1 |
Null value when no check number in historical data with last 6 weeks |
|
|
0 |
High deviation 1000+ |
|
|
|
1 |
Medium deviation 100-1000 | ||
|
2 |
Low deviation 1-100 | ||
| Low Risk |
3 |
Low volume (<=3 items) | |
|
4 |
Multiple ranges | ||
|
5 |
In range |
When the ‘check number range deviation’ category was analyzed, the checks with <= $100 and check numbers with < 100 value were excluded. Just taking mobile check deposits of 2 months of activity from the 3 banks, translated to 2M+ checks. Though the overall return rate was 124 bps but was highly skewed towards high-risk checks and clearly triage the risk into high vs. low-risk checks. After the risk categorization, % of items, return rate and % of return amount were analyzed using various fraud consortium features. It was observed that there is 5X likelihood of fraud in high-risk checks vs. low-risk checks, just based on transaction activity.
First Party Fraud is a growing $100 billion problem. The challenge is that it is much harder to detect than identity theft or synthetic fraud, as the perpetrator is using their own legitimate personal information. Signals like having multiple closed accounts due to fraud or accounts with many authorized users can indicate higher risk but measuring true consumer intent is extremely difficult. To effectively combat first-party fraud, organizations need a comprehensive, cross-industry approach i.e. fraud data consortium.
Using AI-powered analytics and machine learning to identify predictive patterns of fraud behavior can help both in account-level, transaction monitoring and several lending related use cases to intelligently detect fraudsters.
- https://www.federalreserve.gov/paymentsystems/fr-payments-study.htm