Fighting voice fraud with big data analytics

Carriers lose an estimated $40 billion per year to fraud, according to the Communications Fraud Control Association.

In the voice market where profit margins are extremely challenging, it is critical that profits are protected and minutes are monetised. For some carriers, fraud loss might just be a small percentage of revenue. However for some others, it can lead to stagnating profits and even company insolvency.

Fraud loss not only hurts the carriers’ balance sheets, but also damages their relationships with customers and long-term trust in the industry.

Carriers must take a proactive approach to fight against fraud and find ways to prevent and mitigate continually changing threats. No one believes in a ‘wait and see’ approach anymore as the risks are too great. A proactive approach to fraud is critical and that is what our R&D team in Silicon Valley has been working on.

Cataleya’s strengths has all along been about real-time network visibility and big data analytics. The team has developed ways of using such network data and analytics captured from our next generation session border controller (SBC) Orchid One to identify and mitigate fraud, on top of guaranteeing quality of service (QoS) and experience (QoE) on the network.

Orchid One is able to capture network data and analyse it in real-time by monitoring the traffic flow on the network and how the sessions are performing. As most of the current SBCs capture only the historic data like the call data records (CDRs), Orchid One differentiates itself by being able to match historic data with real-time analytics to deliver a new level of network visibility capable of identifying fraudulent activity and increasing the ability to prevent fraud. Not only that, our user friendly web-based dashboard that comes with the Orchid One platform, also provides comprehensive visual reports of the session behaviours which makes identifying of fraud patterns easier.

Acknowledging the industry challenge that fraud is continually changing and difficult to be kept up to date, we have further incorporated the machine learning algorithms into the Orchid One, which requires minimal user input and is globally scalable. The machine learning algorithms in the Orchid One platform monitors the network patterns and once it detects anything unusual, the Orchid One will trigger alerts to the carriers or operators to take necessary action. Fraud scenarios are then automatically logged to be used for preventing similar fraud cases in future. We are continually updating a global database of fraud threats that evolves as quickly as the threats themselves. This allows us to remain agile and move the fight against fraud beyond rule-based approaches into real-time machine learning. Other than identifying known methods of fraud, now we can also predict unknown or emerging threats.

In addition, we have developed the deep voice analysis to where it can actually examine signalling and media in real-time and monitor call quality, set up and disconnect times, as well as call behaviours. This is incorporated into both our CDRs as well as part of our machine learning algorithms. Such comprehensive analytics can be obtained because Orchid One, being a SBC, is deployed right in the call path and all calls have to pass through several levels of scrutiny from our machine learning algorithms to ensure they are not fraudulent. Not only that it is a very powerful self-learning platform, it can make more accurate predictions about the fraudulent activities with no noticeable delays.

Every day, we are innovating ways to enhance both network efficiency and profitability into a single solution. We are pushing the SBCs into a whole new realm and this will benefit the carriers and operators that deploys our platform in their network.