Can banks outsmart AI-driven fraud before it’s too late?

Between April and September 2024, 18,461 frauds were reported, involving a staggering ₹21,367 crore—an eightfold jump compared to ₹2,623 crore last year. 

The number of cases, too, has surged by nearly 28%. Beneath these numbers lies a growing threat that financial institutions are struggling to counter: sophisticated fraud in loan applications.

Digital imposters

Identity fraud is at the heart of this crisis, with fraudsters pretending to be genuine applicants to get loans. Digital onboarding has improved with Aadhaar-based OTP verification and face-matching technology, strengthening security. However, fraudsters are adapting just as fast.

Today, digital applicants must verify their presence in real-time by matching a live selfie with their Aadhaar photo. Yet, as financial institutions tighten security, fraudsters continue to innovate, finding new ways to bypass even the most advanced safeguards.

Also read: India’s KYC crisis: A bank dedicated to welfare payments could resolve it

Deep fakes, powered by artificial intelligence, are now being used to bypass even the most advanced verification systems. These hyper-realistic videos and images mimic genuine applicants, deceiving digital platforms designed to authenticate identities. 

Synthetic identities, fabricated profiles built by blending real and fake information, are also on the rise. Last but not the least, mule accounts, are increasingly being used to launder the proceeds of financial crime.

Offline onboarding processes, often devoid of these advanced fraud checks, exacerbate the problem at times. Fraudsters target institutions that rely on less rigorous manual verifications. Alternatively, they collude with employees and officials to circumvent controls.

When pay slips lie

The problem doesn’t stop at identity fraud. 

Income and employment fraud are wreaking havoc on lenders’ ability to assess creditworthiness. Fabricated salary slips and manipulated bank statements have become common tools for fraudsters seeking to inflate their financial profiles. 

Bank statements, one of the primary sources for verifying income, are being doctored to show false salary trails. Fraudsters often deposit fake salaries into accounts and quickly withdraw the funds to create an illusion of legitimate income.

While the account aggregator framework has simplified income verification by providing real-time access to bank account data, it is still in nascent stages of adoption. 

Also read: RBI’s switch of a domain name for banks won’t really help tackle online fraud

Furthermore, detecting fraudulent salary trails requires sophisticated behavioural analytics to identify red flags, such as high debits soon after salary credits. AI-driven pattern recognition and anomaly detection are the need of the hour.

Over the last couple of years, sophisticated fraudsters have been setting up or acquiring access to shell companies to fabricate work histories and employment records. These fake employers issue forged salary slips and offer letters, often passing superficial verification checks. 

Financial institutions traditionally relied on the Employees’ Provident Fund Organization (EPFO) database to verify employment, but this method is far from comprehensive. 

Small businesses with fewer than 20 employees are not required to register with EPFO, leaving a significant portion of the workforce outside its purview.

The absence of a centralised database for verifying non-EPFO employers and employees creates a fertile ground for fraud. While data for MCA-listed entities is readily available, verifying smaller, unregistered businesses requires lenders to dig deeper. 

Authenticating an employer’s operational history, revenue streams, and expenditures has become essential, yet remains challenging without reliable data sources. The lack of a government-backed repository for these businesses is a glaring gap in the system.

The consequences of these frauds are far-reaching. Financial institutions face direct monetary losses and risk eroding trust in the lending ecosystem. As fraudsters grow bolder and more resourceful, the fight against loan application fraud has become a race against time.

The digital shield

Companies are now actively leveraging advanced machine learning models to detect tampered documents and AI-generated selfies, making it a critical ally in combating identity fraud. 

Artificial intelligence and machine learning solutions go beyond surface-level checks, offering deep due diligence that encompasses business verification, income validation, and active operations assessment, effectively transforming fraud detection by uncovering patterns that traditional methods often miss. 

AI-driven computer vision models have revolutionised tampering detection in document verification. These programs examine documents pixel by pixel, spotting minute changes like mismatched fonts, altered signatures, or altered photos that might go unnoticed by humans.

Also read: Vivek Kaul: Financial frauds evolve fast but it’s not as if we’re helpless

AI is a vital weapon in the battle against financial crime because of its capacity to identify, evaluate, and adjust to changing fraud tendencies, and it is critical for financial institutions to take the tech-first approach to protect themselves and their customers from evolving threats.

Ashok Hariharan, CEO and co-founder of IDfy, an Indian identity verification company.