ASC BLOG

Fraud Detection: Identifying potential cases of fraud and compliance violations

There are many types of fraud - whether on the phone, via internet or using mobile devices. As technology advances, fraudulent techniques are becoming increasingly elaborate - fortunately at the same time solutions for fraud detection are also becoming faster and more precise. Darien Šobar, Vice President Business Development at ASC Technologies has some valuable insights on how to identify and prevent potential cases of fraud and compliance violations.

First of all, what is fraud detection?

Fraud Detection prevents, detects and responds appropriately to fraudulent acts in companies. This requires special analysis systems which:
  • Capture and transcribe call data,
  • Spot dubious behavior of customers or employees and flag it,
  • Search for keywords or phrases that have been flagged as potentially fraudulent, and then generate a report for an administrator to further review.
An automated pre-selection of potential cases of fraud helps financial institutions save time by requiring them to focus on a greatly reduced number of conversations. Fraud detection reduces the number of potential cases or at least the time needed to detect fraud through automation. Today, solutions based on artificial intelligence are increasingly offering support when dealing with fraud.


How do fraudsters proceed?

Criminals are increasingly trying to get hold of confidential information from companies over a wide range of communication media. This is especially the case in unfamiliar situations such as working from home which makes it difficult for employees to distinguish between legitimate requests and attempted fraud. There are often no established processes in a company and there are no colleagues on site who could possibly intervene if they notice something unusual.

Hackers use this opportunity to get at user data. To do this, they collect as much information as possible about the company and their target person in advance to build trust and get employees to release confidential information. It does not necessarily stop at individual attacks, but rather fraudsters exchange scripts among themselves that contain information, for instance when something is checked during a telephone call and how the target person can best be approached. Advanced voice recognition can help to detect such fraud attempts by checking certain calls for certain phrases.

Keywords and phrases are also defined by regulators or compliance officers within an organization, such as a bank or an insurance company. If you are a member of the financial services industry, it is recommended that you follow the guidelines and the regulatory directives of MiFID II or Dodd Frank, depending on whether or not you’re based in the US or the EU. These compliance guidelines provide companies with directions regarding what information must be recorded and what is considered fraudulent behavior.


How can companies act against fraud?

Companies that want to do something to prevent or detect fraud must first learn about the threats they face. This includes an analysis of how much possible fraud would cost the company and how many cases there could possibly be in a given period. The respective industry of a company is an important part of this. From a financial point of view, a bank has much more to fear from fraud than, for example, a telephone company. But the losses a company can suffer through fraud are often underestimated. With the proper software solutions, however, it is possible to capture the full extent of the potential losses in order to maintain an overview.

Companies who know what type of fraud they are most exposed to can take better fraud prevention measures. This can involve offline or real-time fraud detection. Offline solutions require less integration and development effort but are not as powerful as real-time solutions with regard to certain fraud methods. Real-time solutions quickly capture, transcribe and search the call and then send an alert to a third party within the organization responsible for the fraud. Depending on the individual company's risk profile, both types of fraud detection can be used.

The first steps include analyzing the situation of the threat and then deciding on an offline or real-time solution. Then you have to do blacklist management. Blacklist management is the crucial point when it comes to fraud detection. Blacklists consist of people who have already appeared as fraudulent. Most companies maintain a variety of lists also including watchlists and whitelists. Watchlists are lists of potential customers at risk or suspicious persons. Whitelists contain customers who cause fraud reports but have been shown to be legitimate.

In this context, speech recognition can also help to optimize fraud prevention: Advanced voice recognition gives you a variety of ways to define rules that help you track down fraud. The three basic speech recognition approaches, keyword and phrase spotting, phonetics and LVCSR (large-vocabulary continuous speech recognition), differ in the type and depth of analysis, the effort needed for preparation of the system, and the startup costs.

But fraud detection is not only a matter of the software, but also of the training of the agents, bankers and traders who handle the conversation. There are tools that enable employers to train their contact center employees and other consultants on what they can and cannot say during a call.


In which industries can fraud detection tools help to prevent fraud?

Looking beyond the financial sector, there are fraud detection use cases in companies or public health organizations. For example, in healthcare, when customers need to provide health care information. These records must be kept in accordance with established guidelines regarding GDPR and the right to privacy. At the same time, there is always a way to create a log trail when fraud has been committed with customer information, which a company can then quickly identify and act accordingly. But in fact, any company that does business and shares information, must have a fraud detection solution to protect the customer and themselves.


Can you give us an actual example of the use of compliance software to prevent fraud?

We recently had a customer from the financial sector who wanted the following requirements to be met:
  • The brokers and traders had to work from their home office and communicate via Microsoft Teams. This of course had to comply with regulations such as GDPR and MiFID II that require that client conversations and transactions are recorded.
  • They needed a daily report on the total number of calls made by brokers and traders.
  • This report also had to record how many recordings the individual compliance officers listened to on that day. With the requirement that the compliance officers only had access to the recordings for which they were assigned.
  • It should also be ensured that if the compliance officers think that fraudulent activities are taking place, they take appropriate action and report the fraud to their superiors. In addition, those compliance officers must also be monitored to ensure that they do not take advantage of the information they have heard.  
To meet the customer requirements, we delivered ASC Recording Insights a native compliance recording application for Microsoft Teams. With ASC Recording Insights the customer was able to
  • record all transactions and conversations of employees from home but also from any other place using Microsoft Teams
  • generate reports on the number of calls made daily
  • define rules and access rights for individual employees so that compliance officers can only monitor the calls assigned to them
  • monitor the activities of compliance officers and see in a report which records they have examined
For communication that is not taking place via Microsoft Teams but with other communication channels there are solutions such as ASC INSPIRATIONneo that can help prevent fraud.


Looking ahead, how can Artificial Intelligence improve fraud detection now and, in the future, what will be possible?

With advanced speech recognition, you have a wide range of options to define rules that detect fraud. So if one is talking about audio transcription, you can quickly identify which words or phrases that are used might indicate fraudulent activity, and if combined with something such as sentiment analysis, where the Artificial Intelligence measures the pitch of a voice and maybe the speed someone is speaking or the volume at which someone is speaking. There are also ways of detecting whether a person is dishonest, which could also be used to mark certain conversations as fraud. And in the future, the algorithms will be able to provide even a higher level of service.

With Artificial Intelligence and recording and analytics solutions you will be able to detect even faster than before, because all the different algorithms that are currently available will improve over time. Even the transcription of different dialects that are spoken will be possible and differences in dialogue will easily be recognized.


Conclusion: Fraud detection is crucial in order to protect any business

Fraud is serious, it is problematic, it harms not only individuals but also organizations. In general, every business potentially exposed to fraud should have a solution for fraud detection. Especially financial institutions need an internal playbook when it comes to monitoring conversations of a financial character. And they should also have someone designated, such as a compliance officer, to monitor employee behavior and, of course, customer behavior. There are various tools for this purpose, e. g. recording and analytics solutions.


Author: Darien Šobar

Darien Šobar is Vice President of Business Development EMEA at ASC Technologies AG. He is responsible for building awareness and driving adoption of ASC’s compliance recording and analytics solutions together with Microsoft and partners; always interested in working with creative minds that are building practical solutions to tomorrow's problems.







2020-07-20
SHARE





Links


Additional
Related Blogs