Financial fraud occurs when a person intentionally takes valuable assets or money from another person via deception or criminal activity, for the purpose of personal gain. There can be various types of frauds wherein the innocent investors are befooled. For example: Ponzi schemes which guarantee high returns for fixed term investments, but rather they are not invested at all. Identity frauds, where someone impersonates you and gains your personal information to steal, or phishing where innocent bank customers divulge their personal details to a mail which they perceive has been sent from an authorized institution. There has also been a sharp rise in credit card frauds like skimming, wherein fraudsters swipe the card through a “wedge”, which records all the information on the magnetic strip; or they counterfeit cards.
The number of such financial frauds saw a rise of 15% y-o-y basis in 2018-19, according to RBI’s annual report. The total amount involved in such fraud also rose by 73.8% in this year. Fraudulent app transactions have increased by 680% between 2015 and 2018. An interesting observation has been, that such rise in frauds is constituted majorly by PSB’s, followed by private sector banks and foreign banks respectively. Number of frauds relating to advances led the list, followed by credit card, internet and deposit related frauds. Also, the minimum amount of such frauds was relating to amount under Rs. 1,00,000 as they constituted only 0.1% of the total cases, the amount extending to as much as Rs. 100 crores or even more.
Thus, the need of the hour is strict fraud monitoring through proper database and information system interlinking, and imposition of heavy penalties along with supervisory action to curb the heinous actions of such perpetrators. The question remains how?
These days, such frauds have become more nuanced and organized, using machine learning algorithms to defraud businesses, so as to become undetectable. So, what’s needed to thwart fraud is exactly the same as the problem, Artificial Intelligence and Machine Learning. According to AI Innovation Playbook, 63.6% of Financial Institutions believe that AI and ML are an effective tool for preventing fraud. We don’t only need to focus on fraud mitigation but also real-time fraud detection.
Machine learning has become such a viable tool for fraud detection because it can deal with large amount of data, from varied sources, and can identify abnormal patterns that humans can’t identify. Thus, machine learning can easily spot any abnormality in consumers’ behavior, and even back the company against insider fraud; as it can use the data from within the organization and identify any non-routine behavior of the employees, like exposing data to outsiders. While this data analysis takes place in seconds, it consumes a lot of time for a human analyst to identify the such complex patterns or irregularities. This increases the quality and efficiency of fraud analysts, as their burden is released to a great extent. Also, as the accuracy is increased, the number of false positives and false negatives decline; saving cost, time and effort. False positives is a situation where correct transactions are treated as frauds; false negatives is a situation where actual threats and thefts are missed.
As the organized crimes today have become highly sophisticated, our one-size-fits-all solution would definitely produce sub-par results. Thus, we require both supervised and unsupervised machine learning to tackle security. Supervised learning examines events and trends from the past i.e. it uses historical data to find patterns not discernable through rules or predictive analysis. While unsupervised learning helps us in finding anomalies and links between emerging factors i.e. it concentrates on the future. By combining both supervised and unsupervised learning, it gains insight from past as well as the future, promising higher level of accuracy and clarity about customers’ behaviors. Thus, a fraud analyst receives a 360-degree view of transactions, by seeing related historical data and detecting anomaly into real-time activity using unsupervised machine learning, which helps him in managing risk well.
Artificial Intelligence also makes it possible to detect frauds in real time than having to wait for six to eight weeks, until a chargeback request is filed with the bank or financial institution. A chargeback request is when a consumer requests the bank to withdraw a particular amount of money from a merchant’s account, as that was the amount of fraud that took place with the consumer. However, AI is able to detect fraud in seconds, which seems as a blessing.
Mastercard was amongst the first financial institutions to have applied AI and ML for fraud management, and reportedly the number of financial frauds reduced by a whopping 80%. AI and ML has helped the firm identify extraordinary consumer behavior, and build a database of specific consumer pattern, which can be leveraged in the future to detect any abnormal behavior indicating theft or financial fraud.
AI is thus causing massive upheaval in the banking and financial services industry, however, it is not going to be devoid of challenges. There exists a dearth of quality data, and diversity in languages and demographic features. Along with this, many businesses are concerned about the time, effort and expense associated with implementation. However, we cannot sideline the fact that AI and ML have saved our society from huge quantum of frauds taking place, which makes our dream of a cyber free country much more achievable.