Online banking fraud detection technology
Online Banking Fraud is rising and as more criminals are using more sophisticated tools, it’s important for banks and payment providers to use fraud detection technology to combat this.
03 December 2021
Online banking fraud is rising and as more criminals are using more sophisticated tools, it’s important for banks and payment providers to use fraud detection technology to combat this.
According to UK Finance, losses from remote banking fraud rose by 67% in H1 2021, compared to the same period in 2020, from £79.7 million to £133.4 million. As a result, financial institutions are investing more time and money to develop fraud detection algorithms.
In this blog we’re going to take a look at
- The importance of online banking fraud detection technology
- Common issues in current online banking fraud detection
- Online fraud detection and machine learning
The importance of online banking fraud detection technology
Fraudsters will continue to innovate their approach to online fraud as online banking continues to be a lucrative target for fraudsters. One of the biggest methods currently to access a victims account is through phishing, an email containing a link or attachment pretending to be from a trusted source so the victim gives away personal and banking details. They are then stolen and used to access their accounts.
Phishing is becoming more sophisticated and harder to spot. In fact, one in every 4,200 emails in the UK is a phishing attempt, according to Symantec. Online Banking is becoming more and more popular, with approximately 4 in 5 of us using it* – so it’s a big target for fraudsters.
Read this article in Digital Wings to find out more about phishing.
In order for fraud detection technology to be effective, it needs data… lots of data. Banks and financial institutions use hundreds of pieces of information to catch fraudsters, analysing and knowing what is usual for the account holder, to help catch unusual transactions.
Processing huge amounts of data can be expensive and labour intensive. More and more financial institutes are investing in big data analysis and machine learning to process their data, including customer’s transactional information. Machine learning algorithms can be used to help spot patterns and unusual behaviour.
* source: https://www.statista.com/statistics/286273/internet-banking-penetration-in-great-britain/
Common issues in current online banking fraud detection
With the rise of fraud and more sophisticated phishing attacks it’s important to stop fraudulent transactions, but there are some big challenges when it comes to detecting fraud.
False positives
A false positive is when a real transaction is flagged as suspicious, often resulting in cancelling the payment or locking the real account holder out of the account. When working with large amounts of data and large amounts of transactions it’s really important to stop fraudulent transactions in the moment, but false positives can:
- Cause reputational damage to the financial institution
- Have an impact on a legitimate customer’s finances
Reducing false positives is an important step to the future of fraud detection and this can be done by automating fraud prevention technology. By using real-time data, banks can develop profiles on customers which can assist in flagging transactions that are usual for the specific customer, rather than a blanket approach.
Difficulty in scaling
Scaling fraud detection technology tools to keep up with increasing levels of transactions can also be a challenge – but not just with fraud detection. Artificial Intelligence systems and deep learning machines need to deal with real time data and make instant decisions, as well as constantly developing it’s algorithms and ‘learning’ from outcomes. There are 2 main difficulties when it comes to scaling the technology to a large scale bank:
Technical performance
Running big data analysis requires a lot of computer processing and processing of large amounts of data. In order to ensure that false positives are reduced and fraud detection is effective, lots of different data points are needed – meaning millions of data points will be processed.
A system that flags possible fraudulent transactions using real-time data needs to be able to process this instantly – if the computer is unable to get through the live data for a day this won’t be effective when it comes to stopping fraud.
Amount of data
As well as the processing of the data, this needs to be stored. Whenever you use data in a different part of an IT environment, cyber security will need to be high. By feeding confidential data into an AI system, there’s another area of the environment that could be a ‘point of breach’ when it comes to a hack, so the actually IT infrastructure will need to be well thought out and planned.
Online fraud detection and machine learning
So, what’s the difference using machine learning? A lot of companies around the world still use rule-based systems to detect fraud. If certain rules are met, something is done. For example, transactions over a certain limit or paid to a certain country might be flagged for someone to review it. This can be an easy way to stop a lot of fraudulent transactions when you already know patterns of fraud, but as fraudsters change their tactics and change their patterns, you can’t rely on rules alone.
How does machine learning work?
There are four main steps to machine learning:
- Existing data of both genuine and fraudulent transactions and data about the account is fed into the model, allowing the computer to detect patterns based on previous data
- Using this model, it will give a probability of fraud and an explanation to the analyst on how that probability is met
- The outcome on whether it was fraud or not is fed back into the computer, ‘learning’ whether the prediction was true or not
- Once predictions are at an acceptable level, the data analysts can set rules on what to do with the account (i.e. lock it or stop a payment) once a certain probability has been met.
Watch the video below from Amazon about their fraud detector using machine learning:
In short, there are still challenges when it comes to preventing fraud while keeping false positives low to lessen the impact on genuine transactions, so it’s worth making sure you’re familiar with some of the ways you can keep your online banking details safe.
To find out more about keeping yourself safe, take a look at our digital safety module.
|
To find out more about artificial intelligence, take a look at our AI module.
|
