The banking sector is steadily moving towards digitisation, with artificial intelligence and metaverse predicted to drive commercial activities in the coming years. However, another demon is slowly raising its ugly head in the form of fraud.
The incidents of online fraud attacks are rising at a good speed. As per a PwC crime survey, in 2021 alone, the global banking sector suffered some USD 42 billion in losses at the hands of scammers.
Outseer data, which came out in the first quarter of the 2022-23 financial year, says that APP fraud (where consumers are duped into romance and crypto scams), made up some 75% of all fraudulent banking payments. While customer education on these matters is important to stop fraud from happening, banks’ role becomes even more important, when it comes to analysing and countering these threat actors. That’s where the participation of fraud analytics becomes important.
Understanding fraud analytics
Fraud analytics performs a set of functions such as data mining, machine learning, and putting into use applied data science to detect and prevent fraudulent transactions from happening.
The ecosystem involves banks accessing information from a large dataset and then analysing and examining those details to establish fraud patterns, following which they come up with their threat response mechanisms. To make it highly successful, the key word here is more and more data.
The efficiency of a financial institution to identify and prevent fraud early depends on how good they are at sharing data with businesses and law enforcement agencies across time zones and geographies. With the power of machine learning and artificial intelligence, automation can immediately take corrective action when these threats are detected.
Pros of fraud analytics
Quick threat perception time: All the mechanism needs is the combination of machine learning and artificial intelligence to deliver effective results. The fewer fraudulent transactions are, the better it is for the banks, in terms of lessening financial losses. Deploying artificial intelligence to do this job gets rid of the monotonous job of manual data research. With effective artificial intelligence, financial organisations can be a step ahead of the threat actors by having a fast, smooth investigation and research process. The more you identify the scams early, the more you can save your banking data, customers’ money and most importantly, your intellectual property.
Not only money, but protects brand values as well: We have come across many reported instances of fraudsters impersonating bank officials and tricking their victims into revealing their sensitive financial details such as bank account numbers, credit and debit card information. These scams not only hurt the customers, but destroy the banks’ brand values as well. As per UK Finance data, in the first six months of 2020 alone, banks saw some 15,000 impersonation scams, which resulted in a total theft of £58 million.
What’s even scarier is that during the same year, a Hong Kong bank manager got duped by scamsters who used artificial intelligence voice cloning to pull off a USD 35 million heist. The UAE investigators had to take the help of their US counterparts to trace the stolen money which got transferred into the accounts of Centennial Bank, whose director’s voice was cloned by the scammers with the help of “deep voice” tech. The fraud involved 17 anonymous cyber criminals in total.
Spoofed domains and brand smartphone apps, fake social media accounts and phishing sites have become principal weapons for these scammers, which a proper fraud analytics mechanism can detect and remove, with artificial intelligence and machine learning help.
Protecting customers from the dark net grasp
Since the year 2007, some 200 cyber incidents have taken place, targeting financial institutions. In 2017, the G20 warned that these incidents could “undermine the security and confidence and endanger financial stability.”
With the rapid growth of the internet and cashless economy, people are making their transactions through digital payment portals and credit card scanners. Here comes the threat from darknet operators, as they can hack one of these cyber tools and hijack the crucial banking details of the customers. Using the compromised data, these criminals can log into victims’ accounts and perform malicious acts such as changing the account security settings and making purchases.
The banks have to go through the traumatising process of shelling out a huge sum to release the data. Recently, a report from IT firm Claysys has warned about a new phenomenon called “spoofing”, where these darknet operators are opening fake websites while imitating the original URL of the bank. If someone is using his/her sensitive information and passwords to access these websites, their data is directly landing in the laps of these threat actors. Even third-party tools such as chatbots and customer relationship management software used by banks to communicate with their customers can be hacked as well, as per the above report.
Using artificial intelligence and modern data science, fraud analytics mechanisms can take care of the above threats by accessing and analysing the customers’ transaction data from his/her login ID. This process studies the “suspicious money transfers” and classifies the destination accounts as “friends” or “foes”. Accounts identified as “foes” will come under the scanner of the banks and cyber security officials.
Banks need to invest heavily in machine learning
Already financial institutions are using a large chunk of their operational costs on maintaining their fraud detection software. Same thing they need to do on the data science front as well. The daily operations of a bank revolve around a pile of data. While gathering and maintaining consumers’ data can be performed with regular tech solutions, we are living in the age of cybercrime, with dark net operators always up on their toes to find new victims and swindle money out of them.
How can a bank identify someone logging into one of its customer’s online accounts as actually a genuine user or a dark net operator?
As per cybersecurity research firm “One Span”, there are more than 15 billion stolen credentials up for sale on the dark web. Cybercriminals can purchase them for an average of USD 15.43, when it comes to abusing stolen customer credentials. For big financial institutions, the price tag is over USD 3,139.
To stop this, investing heavily in machine learning is the only option. Machine learning is all about coming up with intelligent algorithms. Apart from analysing huge data flow with the help of complex algorithms to figure out potential fraud patterns, machine learning can also help computing devices to respond to such threat situations. These algorithms take the help of data sets to identify fraudsters and genuine customers. These programmes gather and categorise data all the time. They have two catalogs, “good data”, information about legitimate customers and “bad data”, information dealing with fraudulent transactions committed by potential scamsters.
Under machine learning, the algorithms are “trained” on differentiating between genuine customers and offenders, with the help of massive data sets. Once the process gets completed, the programme becomes “business specific”, ready to be embedded inside the bank’s fraud detection system. To ensure the success of this solution, these algorithms needed to be updated on a timely basis.
These algorithms can not only process big sets of data quickly, but they can also perform repetitive tasks and detect even the smallest of changes in data patterns. This function helps the bank to identify potential frauds much earlier than the time taken by a team of human analysts to perform the same work. By analysing hundreds of thousands of payments per second, it saves the organisation’s costs, while improving its response mechanisms, in case of financial irregularity.
Since these algorithms continuously collect and analyse new data, the accuracy factor against fraud always remains high. As opposed to human analysts bogging down, while finding and evaluating the relevant information, amid a huge data flow, these machine learning solutions are time, cost savers for businesses, while providing top-notch efficiency and accuracy against threat actors.
These algorithms work on the principle of “the more data flow gets bigger, the better it becomes”. With the ability to identify the smallest of the changes in the data flow patterns, these tools cut down the rate of undetected fraud.
A recent report from “Telus International” laid out a roadmap for financial institutions, in terms of making the best use of fraud analytics software.
The report said that the banks can use machine learning to not only create catalog of financial and non-financial types of data, but to create customer profiles as well. These profiles will help these institutions to understand and predict customer behaviors. A single profile can be updated after each transaction, following which the artificial intelligence-guided mechanism can decide upon the transaction pattern and issues red flags if there is something fishy.
These transactions can also be given a fraud score by the artificial intelligence, with the help of data carrying the content from past legitimate transactions. This will help the banks to decide their action plan based on their fraud and financial risk parameters. This fraud score can be decided by using variables such as transaction amount, transaction times, frequency of credit and debit card usage, and many more. These scores can be programmed for purposes such as automatically approving a transaction, sending it for a threat review or rejecting the transaction request.
Since this particular fraud analysis module will be using machine learning, it can even use neural networks to have the advantage of making quick decisions. This solution will be flagging the massive number of fishy transactions, which otherwise would have been a difficult function to perform for a team of human analysts. On top of that, artificial intelligence will give periodical lists of flagged transactions, which requires more investigation. Here, the role of Augmented Intelligence will play a crucial role to prioritize transactions, which require immediate probes from the banks.
Even the KYC (Know Your Customer) processes can be introduced to AI-based solutions, as they would be able to verify customer IDs and banking documents with methods such as fingerprint matching and facial recognition.
Leading strategic consulting and market research firm BlueWeave Consulting in its recent study, has estimated the size of the global fraud detection and prevention market at USD 28.36 billion. The survey also expects the market to grow by 22% and become a USD 110.17 billion sector by 2028.
The 2020 incident of the Hong Kong banking professional getting duped by dark net operators through artificial intelligence voice cloning is a grim reminder of how these cyber criminals are finding technological breakthroughs to hurt their victims. It’s high time that global financial networks embrace fraud analytics software to get the much-needed upper hand over these scammers.