StashFin Uses AI and ML for Fraud Detection

Financial frauds are growing increasingly complex day by day, and the financial service sector cannot afford to take them lightly, be it fintech companies, retail banking, or insurance service companies. These financial frauds are taking place at a scale that puts at risk the whole industry. In 2020-21 alone, the RBI financial institutions have reported frauds worth Rs. 1.38 trillion, this data excludes undetected frauds and frauds below Rs. 1,00,000. It is becoming impossible to detect them through manual processes.

According to the Boston Consulting Group report, the valuation of India’s fintech industry is said to reach $150-160 billion by 2025. The automated processes of AI and Machine learning have not only resulted in reducing the number of cases but also aid in detecting frauds before they could occur. Hence, Fintech companies are also adapting to machine learning algorithms to keep pace with complex frauds.

StashFin Fraud Detection

StashFin the digital money lending company provides a credit line card powered by VISA or Mastercard. The interest rates are charged based on credit score, which will be reduced as the borrower’s credit profile improves. The onboarding process for StashFin involves a 5 Step procedure, wherein the customers have to share their Aadhar and PAN card pictures.

Then, with the help of a machine learning model, details are extracted to fill the registration form. Once the OTP is entered by the user, the underwriting model:

The system goes through 3,000 variables like the user’s credit score, amount of credit applied for, banking history, etc. And with the help of standard APIs and basic models, StashFin verifies the validity of the uploaded documents. It also uses the histogram algorithm to detect if the selfie uploaded is of a real person or a synthetic.

AI (Artificial Intelligence) and ML (Machine Learning) Models for Risk Positioning

StashFin receives around 150 to 200 loan requests in an hour, these loans are processed with the help of AI and ML models. These models also send reminders by studying a user’s repayment habits instead of calling them frequently. In certain cases, the fraud detection models can reject a good customer due to false positives. So, to ensure that this doesn’t happen, StashFin rechecks the rejected application manually with human intervention.

StashFin currently has about a million customers on board, it aims to increase this number by 7 times in the next 24 months. Hence with the help of their in-house engineering and data scientists’ team, StashFin is striving towards improving their product, as well as, their fraud detection systems as they scale up.