We are in the midst of an economic cybercrime wave. According to a PWC report, 51% of organisations had experienced fraud in the past two years, the highest level in 20 years. Of those, 61% experienced external impact, such as damage to their reputation, loss of customer trust, or regulatory action.
Perhaps unsurprisingly, the report highlights how financial services companies and tech businesses are more likely than other sectors to be the victims of fraud. And embedded finance is no different.
Why? For the same reason that it is growing in popularity generally:
- A relatively new financial service lacking regulatory control, standards, or oversight
- Rapid, exponential growth
- A convenient, easy way to move money
As non-finance companies increase their deployment of embedded finance, they will also need to increase the protection they have in place against bad actors who could target them. After all, these businesses will likely have less experience in combating fraud and financial attacks that banks, insurance companies and other financial service providers deal with daily.
Nima Montazeri is the Chief Product Officer of Liberis.
Opening up while locking down
Fraud is a threat that users of embedded finance and its providers need to wake up to, especially when they are looking to offer an exemplary experience of using embedded finance. Cyber defenses that lock down everything, limit the movement of data and demand constant authentication must be delivered in a way that is cohesive with the easy-to-use and frictionless approach most customers expect when it comes to embedded finance.
Integrating this level of security requires more innovative technologies, of which many neobanks and incumbents have already implemented with the likes of 2-Factor Authentication and facial recognition capabilities. When it comes to fraud, however, it is a difficult challenge to overcome. Why? Because only by analyzing data and spotting anomalies can organizations act to prevent fraud. Yet most traditional detection techniques aren’t sophisticated enough for the latest cyber-attacks. Manual and time-consuming, they struggle to analyze even a small percentage of the data that needs to be processed to help prevent these threats.
Powering fraud prevention with AI
So, how do embedded finance providers and their customers combat this? Through artificial intelligence. AI-powered fraud detection and prevention tools can conduct real-time analysis of huge data sets to identify anomalies, flag potentially fraudulent transactions and make it easier for fraud departments to act decisively and, most importantly, rapidly.
But it isn’t just about speed. AI gets better as it is used more, so as it is deployed on more data sets, it will become more efficient at distinguishing between different behaviors and determining what’s legitimate and what needs reviewing.
The solution’s challenges
Of course, deploying AI fraud prevention tools effectively requires care and consideration. Several challenges need to be addressed in order to ensure success
- Combatting a lack of data: AI tools live on data; as noted, the more data, the better they can be. That’s great for big enterprises, but harvesting the volumes of data required to create a baseline understanding for smaller organizations could be difficult.
- Understanding ethics: Human understanding of AI can be limited, so there can be a struggle to explain how it generates its results. An input turns into an output without a clear idea of what has happened, creating a ‘black box’ effect that can make it harder to be transparent on processes and, therefore, fair and accountable for outputs.
- Keeping data private and secure: Large amounts of data increase the risks of breaches and raise concerns around the privacy of personal information. As such, security and responsible use of data must be a priority.
- Constant compliance: Laws like the European Union’s General Data Protection Regulation (GDPR) ensure that the use of personal data follows strict rules, with significant sanctions for companies that fail to comply.
So, any organization deploying AI tools needs to be conscious of these challenges and have a clear plan to overcome them to allow the solutions to be used effectively.
AI in action
When they do, the results can be significant. For instance, the major Danish financial institution Danske Bank must manage a large attack surface as almost all of its customer interactions occur digitally. Historically, the bank’s fraud detection rate was 40%, with 1200 false positives daily. This meant that not only was it missing opportunities to prevent fraud, but much of its resources were spent on false alarms.
To combat this, the bank applied AI-powered analytics to improve its fraud detection process while reducing false positives. It certainly paid off: a 50% increase in true positives and a 60% reduction in false positives, aiming to reach as high as 80%.
Making embedded finance safer in real-time
As attacks become more sophisticated, the tools used to combat them have to become equally advanced. Embedded finance has much to offer customers and businesses, but those benefits make it an attractive target for bad actors. Using AI to identify fraudulent behavior can help providers and users of embedded finance identify criminal activity and act swiftly to prevent customers from losing out.
