Fighting E-commerce Fraud with Technology
In May 2017, Amazon lost INR 69.91 lakh to e-commerce fraud in Bengaluru, and Flipkart was cheated off of INR 1.05 Crore in Kota. These frauds happen across three categories: buyer side, merchant side, and security system side. The methods are ingenuous and cover a broad range including chargebacks, selling counterfeit items, account takeovers, and identity theft.
According to reports, in India, the highest percent of e-commerce fraud happens through Return to Origin (RTO). Buyers abuse return policies and grievance mechanisms to claim false reimbursements. For instance, two youngsters falsely claimed to have received empty boxes from Flipkart in response to their order of mobile phones. Their fraud scheme ran for more than a year during which they robbed off mobile phones worth INR 1.05 Crore.
From return to origin (RTO) and friendly chargeback frauds perpetrated by buyers, to mobile frauds carried out via apps, e-commerce fraud is nuanced with many layers and forms. Combating it requires intervention both at the level of policy and technology.
The central Government has taken light of such unfair trade ethics in e-commerce in the country. On 30th July 2019, it passed the Consumer Protection Bill to prevent unethical trade practices and e-commerce fraud. However, attacks grow more sophisticated in step with evolving technology. The current fraud prevention mechanisms used by e-commerce businesses are manual reviews, and static rule-based systems which cannot rise up to the dynamic nature of e-commerce fraud.
In order to effectively curb e-commerce fraud, it is important to adaptively learn and model transaction and user behaviour patterns. Static rules often lead to false positives affecting genuine customer interactions with the business. Using AI based methods such as Machine Learning, companies can reduce the incidence of fraud by 30%. By understanding topologies of relationships between different entities, and building accurate risk profiles, machine learning based tools prevent fraudulent transactions before they enter the system.