Changes in business models, steep competition make risk management difficult for enterprises: R …
With regards to risk management and fraud control, what are the challenges enterprises are facing today?
There are two sets of frauds—one is occupational fraud, which is caused by employees, and second is external fraud. So from a business perspective, it is called managing fraud, risk and compliance, and from a technology and algorithmic perspective it is known as anomaly detection.
According to Association of Certified Fraud Examiners (ACFE), the cost of fraud to organizations is 5 percent of business turnover. In telecommunication world, the revenue assurance can range from 2 to 15 percent. It also suggests that a typical fraud is identified after 18 months, while 50 percent go undetected, and that’s what makes fraud control such a challenging job.
Managing a fraud and detecting compliance is like searching for a needle in a haystack. Collusive frauds are well-planned and difficult to predict through transactional pattern analysis. Large enterprises are usually concerned around securing intangibles data that carry significantly more value than physical assets.
Why fraud, risk and compliance control need a ‘big data’ makeover?
Risk management function within most enterprises are struck with several challenges. Rapid changes in business models and increased competition have made managing risks a major concern.
The traditional techniques and casual methods of doing data analysis fails to show the big picture. Such techniques could lead to generating false results and leave out major incidents. And that’s where businesses want to gather as many data feeds as possible so as to see the big picture.
Big data techniques allow the platform to address issues listed, by creating the ability to handle data-at-scale and process them in real-time. Machine learning algorithms are used to learn from prior data and identify possible cases that require investigation, without explicit rules being defined. This would make sure that the detection rules are updated with the ever changing environment.
What does it take enterprises to build a smarter mousetrap?
The nature of frauds are so unique and varied, that it becomes very difficult to create a traditional model to identify frauds, so one has to use machine learning techniques to deal with such issues.
The challenges in detecting anomalies lie in connecting the dots proactively. Handling huge volumes of data from multiple sources in a speedy manner and collaborating across teams are few concerns that most enterprises face.
Addressing this challenge requires a change in mindset and willingness to use new technologies. This requires continuous monitoring and high end tools that have self-learning capabilities to keep up with the ever-changing face of fraud.
How does Wipro’s anomaly detection platform, Apollo, helps organizations to address challenges in managing fraud, risk and compliance?
Apollo is basically designed for controlling frauds, mitigating risks and compliance, using big data and machine learning algorithms. It has been deployed to address issues such as accounts fraud, policy compliance, data theft, ethical violations, regulatory compliance, harassment on email, etcetera.
The platform manages detection service, with 350 pre-built models for rapid deployment and ability to customize according to needs. It is a scalable platform, which works end to end, including data sourcing, cleansing, reporting, case management, and also compliant with data privacy.
Over 50 investigators work on the results generated and the feedback from those investigations is incorporated into the algorithms for continuous improvement of the model. Apollo prioritizes identified anomalies for investigation, facilitates early, and detect potential frauds.
Comprehensive and continuous surveillance of all data including checks for threats, results in an improved risk climate for the business. Other benefits include improved process efficiency, expense recovery and ensure savings.
What are the challenges Wipro usually face while designing and implementing of an anomaly detection system?
The key challenge is not a technology problem, but more of a mindset issue. As stake holders have always lived in a stable way, they are not very open to a change in their business model.
It is very difficult for them to move from a world of scientific sampling, intuition and rules, to the world of data and analytics, and thus it becomes very difficult for us to give them enough to convince for such platforms.
The typical approach that we take, is to do a big bang implementation and once the customers see value in such platforms and technology, it becomes easy for us to make them understand the true potential.
With Apollo, one can get all their data analyzed, with real-time response; control risks that are driven by data-centric view of any breakdowns; enhance intelligence through common and consistent view of data and red-flags and do away with duplication and excess payouts.
What is your go-to-market strategy for Apollo?
We have an open source platform, with a whole set of rules, and the ability to host it is either on cloud or behind customer firewalls. Most of our customers prefer on premise implementation, behind their firewalls, to have a full control over their data—which we manage remotely.
As we have a whole bunch of rules, it’s easy for us to do a quick deployment and get results in two to three weeks. In terms of pricing we divide it in three parameters—initial license and implementation fee, monthly maintenance fee, and in some gain share construct where we identify the cases of interest and we get paid a small slice of their savings.
What changes do you expect to see in the near future, with respect to cyber security, fraud detection etcetera?
It is definitely important to identify the process discrepancies and detecting fraud in a business, but a key question comes up, is of choosing the right approach for proactive fraud control. One needs to support the fraud control initiative and set precise expectations around the effort involved and the results are interpreted.
It is an interesting problem, for Wipro, to solve across the whole bunch of domains, is by applying ability to gather data and correlating it at a significantly higher scale than the traditional risk and compliance product. We are looking at correlating between physical access logs and HR access logs, and map on to see suspicious transactions. We also offer the ability to build algorithms, which are sophisticated and learn over time the patterns of abnormal behaviors, frauds and risks.
Source: Changes in business models, steep competition make risk management difficult for enterprises: R …
Via: Google Alert for ML