Subex Fraud Management

At the heart of the Subex Fraud Management solution is a hybrid rule engine that covers detection techniques like thresholds, expressions, and trends. Rule engine comprises of a combination of threshold rules, geographic rules, pattern (sequential) rules, combinatorial rules, ratio/proportion-based rules, negative rules, hotlist based rules, etc. which enable you to monitor advanced threats in the network. Rule Test Facility (RTF) in the solution provides a controlled environment for fraud managers to configure and validate rules, before configuring them on the live network. Intelligent alarm qualification designed based on hidden Markov models in the solution ensures an analyst catches at least 80% of frauds by only having to go through no more than 20% of the alarms on the alarm stack.

The Subex Fraud Management system is integrated with machine learning algorithms to mine historical data from fraudulent behaviors to create supervised and unsupervised models. These models enable telcos to find hidden patterns in data for fraudulent activities, which reduces the time-to-detect fraud significantly. Subex has multiple sets of proprietary ensemble models with the capability of state-of-the-art patented automated hyperparameter tuning, and autonomous model selection based on the properties of data, such as the geography, the nature of fraud committed and more. These models have auto-retrain embedded, which get triggered when the accuracy falls below an acceptable threshold. We leverage modern machine learning and deep learning algorithms like LSTM (Long short-term memory), Generative Adversarial Networks (GANs), Capsule Network (CapsNets), elastic net, etc. to detect and prevent fraud in the telecom network.

A proactive approach to prevent fraud in the digital ecosystem

The signaling security module in Subex fraud management is designed to address the traditional and new-age risks across various digital services. It monitors multiple parts of a telco network starting at the layer 2 and going up to layer 7 to detect fraud attacks. By starting at the lower level of the network stack, you can now detect breaches much earlier. The solution employs a proactive approach using a combination of signatures, heuristics, and machine learning techniques to prevent fraud proactively.

Blockchain, as a technology, brings in strong capabilities which can used to tackle telecom fraud. One of the important use cases where this can be implemented with this cryptographically secure database is real-time threat intelligence database. Subex has partnered with the RAG Wangiri Blockchain Consortium which uses Blockchain technology to gather real-time industry threat intelligence about fraudsters. Through this partnership, Subex Fraud Management aims to provide you with a decentralized and cryptographically secure blockchain ledger of Wangiri fraud-related information. Any new data that is uploaded on the RAG Wangiri database is automatically uploaded on the Subex Fraud Management system. New traffic on the network is monitored against this data and fraud analysts can take appropriate actions.