According to the Communications Fraud Control Association (CFCA), the estimated global fraud loss for communications service providers reached $39.89 billion in 2021. This includes various cases of fraud carried out against either CSPs or their subscribers, such as IRSF, bypass fraud, subscription fraud, roaming fraud and others.
Although global fraud loss has been on the decrease due to the providers’ ability to prevent some of the most common cases and increased regard for cyber security, fraudsters and scammers have also been improving their methods to profit from CSPs and subscribers alike.
To address the ever-changing threats to CSPs and their subscribers, a hybrid approach which combines traditional rule-based methods and state-of-the-art Machine Learning (ML) applications is necessary.
Elitnet’s Onyx Fraud Management System (Onyx FMS) is an innovative hybrid system for combatting fraud in telco networks. Empowered by state-of-the-art Machine Learning and Big Data technologies, the system detects and prevents various types of fraud in real time, reduces the fraud window, and maximizes cost savings.
Onyx FMS consists of three key types of components: data collection components, data processing and analysis components, and reporting and enforcement components.
The system gathers data from various points in the network in real time or near real time, including CDRs/SDRs, SS7/Diameter messages, NRTRDE files, subscriber information, and other types of data. The collected data is then processed using a rich set of methods for fraud detection ranging from rational rules on aggregated counters and signaling pattern recognition to ML-based anomaly detection, traffic and subscriber classification, and real-time stream analysis.
KEY VALUES
SYNERGY OF RULE AND ML-BASED METHODS
The synergy of rule-based and Machine Learning methods allows Onyx to reach next level ML precision while maintaining maximum efficiency. Onyx’s ML-based anomaly detection components also allow reliable detection of previously unknown fraud patterns, schemes, and attack types.
FULL ML MODEL LIFECYCLE
Onyx features full ML model lifecycle including runtime, performance monitoring, feature analysis, as well as continuous automated ML model performance evaluation and retraining, which ensures constantly high recall and precision rates of fraud detection.
POWERFUL ANALYTICS AND DATA DRILLDOWN
Onyx provides a feature-rich data analytics GUI for comprehensive statistics and reporting, convenient drag-and-drop rule configuration, ML model feature management, and powerful tools for investigative analysis of individual fraud cases.
OPEN ARCHITECTURE
Onyx allows operators to address new market challenges by leveraging the platform’s open architecture. It provides convenient tools for rule configuration and well-defined component types and interfaces for implementation of new applications.
DETECTED FRAUD TYPES
INTERNATIONAL REVENUE SHARE FRAUD
By employing multidimensional data aggregations and an ML model, Onyx detects IRSF cases on outbound calls and transit calls by A numbers and B numbers, covering various technical fraud schemes such as PBX hijacking, device malware, roaming frauds, etc.
WANGIRI FRAUD
Onyx detects Wangiri fraud by employing a residual neural network, a state-of-the-art deep learning model. After a Wangiri attack is detected, Onyx’s enforcement engine allows to carry out intelligent blocking using IVR warnings to minimize subscriber callbacks to Wangiri numbers.
INTERCONNECT BYPASS FRAUD
Onyx uses ML-based methods such as SIM card clustering and fraudulent behavior recognition models as well as rule-based filtering to detect various bypass fraud cases such as on-net and off-net SIM boxes, CLI refiling, and OTT bypass.
ROAMING FRAUD
To prevent inbound and outbound roaming fraud, Onyx applies a wide range of roaming traffic rules in real time. Roaming fraud is also detected using IRSF ML models by applying them for roamer MSISDNs.
SUBSCRIPTION FRAUD
To detect subscription fraud cases such as stolen or hacked SIM cards, device malware, etc., Onyx uses sophisticated ML techniques such as subscriber classification for fraud risk evaluation, subscriber behavior anomaly detection, and subscriber clustering.
OTHER FRAUD TYPES
Onyx uses its powerful rule engine for anomalous usage detection as well as ML models for traffic and subscriber anomaly detection. Onyx can be employed to detect a wide range of other well-known and well-defined fraud cases as well as atypical fraud cases without a clear, previously known pattern.
Elitnet is a member of the GSM Association and the Communications Fraud Control Association.
|
|
|
|---|
For more information about Onyx FMS, please download the product datasheet and case studies or contact us using the links provided above.
– Back to top –