The evolving threat landscape: How AI and ML are redefining fraud detection in digital payments

By Sandeep Sethi- VP of Sales- India and South Asia, Wibmo, a PayU company

The world is full of contradictions, and technology might just be the most striking one. It has transformed how we live, work, communicate, shop, and manage everyday tasks. Much of what once seemed impossible has now become second nature, thanks to the technological ingenuity of humankind. While it is fair to celebrate these extraordinary strides, it is equally important to acknowledge that every fast-paced transformation brings along its share of risk.

Technology, for all its promise, can become dangerous when mishandled. The same tap which enables convenience for you could also come from a cybercriminal attempting to breach your financial security.

India has been at the forefront of the Digital Banking globally with over 80% of India’s transactions now Digital first led by UPI. However, with financial transactions shifting to digital platforms, cybercriminals have become more sophisticated in their methods of attack. According to data published by the Government of India, between April 2014 and December 2024, losses from such digital payment frauds totalled a staggering INR 733.26 crore across 63,315 documented cases.

Judicious usage of technology enables secure payments
Today, we can purchase anything under the sun with a single tap. Whether booking your next trip, buying groceries at the e-store, ordering a meal, or shopping online from the comfort of your couch, digital payments have made these transactions incredibly convenient. The payments industry is set to undergo significant growth over the next five years, with digital payments arising as a key frontier. India is well-positioned to lead the industry with its Artificial Intelligence (AI) capabilities within the payments ecosystem.

Financial Service Providers (FSPs) in the country have collectively invested over USD 3.2 billion in developing AI and Machine Learning (ML)-powered solutions to improve the accessibility, usability, and overall quality of payment services.

While AI is being applied across the entire payment lifecycle, from initiation to confirmation, our discussion focuses on its role in strengthening payment security.

Why AI-driven payment security systems are the best defense against fraud
Picture this: You have been waiting for your favorite band to announce a concert, and the tickets are about to go live. You are online, refreshing the booking webpage and hoping to grab your tickets before they sell out. The second they are available, you hit purchase, prompting your bank’s systems into action behind the scenes. However, here is something most people do not think about: instances like these are prime targets for digital payment fraud.

So, how is your transaction protected? How does your money stay safe in a world full of phishing links and dodgy websites? This is where advanced technologies like AI and ML silently step in to add additional layers of security to the digital payment infrastructure.

Real-time monitoring: Securing transactions as they happen
So this is what happens behind the scenes: The second you confirm your ticket’s purchase, AI-driven systems begin evaluating that transaction in real time. Any suspicious activity, including irregular purchase patterns and unidentified device usage, can trigger instant alerts or blocks. These systems act immediately, preventing fraud before it occurs, without slowing down the user experience.

Pattern recognition: Identifying subtle anomalies
If you are someone who usually makes low-value purchases, and, contrarily, you buy six VIP concert tickets worth INR 7,000 each. An AI-powered payment security system would naturally flag this activity as unusual. Thus, even if this is a one-time transaction, it contradicts your past purchase patterns. AI models analyse your historical behavior and purchasing patterns to detect anomalies, connecting seemingly unrelated data points to flag complex fraud attempts that traditional systems might overlook.

Adaptability: Staying ahead of evolving threats
Techniques used to carry out payment frauds are changing time and again – whether it is cloned ticketing websites, tampered QR codes, or payment gateways linked to suspicious third-party sites, Machine Learning algorithms, a core component of AI, power the latter’s ability to learn and adapt incessantly. With every new data point, behaviour or information, the system improves its understanding of fraud patterns, adapting in real time without requiring manual reprogramming.

Scalability: Managing high volumes with consistency
When tickets for a major concert go live, millions of transactions are bound to occur within minutes. AI-powered systems are built to scale effortlessly, analysing vast amounts of payment data in real time. This ensures that every transaction is evaluated without delay or compromise in accuracy, even during peak traffic.

Minimised false positives: Accuracy without disruption
When a genuine transaction gets flagged as fraud, it results in inconvenience. Since AI-driven systems consider a broader range of variables, like location, timing, device usage, and spending behavior, they can distinguish between legitimate and suspicious transactions with accuracy. This reduces false positives, promising a smoother user experience coupled with tighter security.

End words
In FY 2023–24, India recorded over 29,000 high-value cyber-fraud cases, resulting in digital payment losses of approximately ₹1,400 crore. High-value fraud cases surged more than four-fold compared to the previous year. To combat this, a federated AI/ML-based fraud detection system has been piloted across key players in the payments ecosystem. This system uses a combination of device intelligence, transaction profiling, and customer behavior analytics to flag suspicious transactions in real time. Despite over 18.3 billion UPI transactions in March 2025 alone—worth ₹24 lakh crore—fraud was limited to about one in every 6.5 lakh transactions. Additionally, AI-driven risk engines analyzing 50–100+ parameters per transaction have led to up to 50% reduction in fraud losses and a 25–60% drop in false alerts.

There is a paradoxical nature to technology. However, even among fears of misusage, advanced technology is not a destabiliser. As digital payments become more central to everyday experiences, the role of AI and ML in fraud detection is no longer optional but foundational. Their ability to act instantly, learn continuously, and operate at scale is what keeps your transactions safe and secure.