Use Cases Include: Robocall / Nuisance Call Detection and Revenue Share Fraud (IRSF and Wangiri)
Beyond traditional types of voice fraud, such as Revenue Share Fraud (IRSF) and Wangiri (missed call) fraud, fraudsters have begun to leverage SIP architectures for fake call centers and robocalls which has led to major damage for mobile subscribers.
Fraudulent activity has caused a dramatic drop in legal enterprise voice traffic, leading to multiple regulator initiatives on preventing Caller ID spoofing and robocalling. This problem is not limited to any geography and is slowly spreading across Europe, Asia, and the Middle East, as regulators prepare to step in and force Communications Service Providers (CSPs) to take action.
Mavenir has introduced the Mavenir CallShield solution to address growing challenges with Mobile Voice Communication Services. Mavenir’s CallShield leverages the Mavenir Fraud and Security Suite framework.
With years of AI/Real-Time Machine Learning (ML) and live call screening capabilities to identify malicious call attempts, CallShield provides CSPs with controls to minimize voice fraud damage, protect subscribers, and revenue.
Protection from Robocalls, Nuisance Calls, IRSF, Wangri and Fraudulent Call Centers
CSPs can no longer rely on using only rules and thresholds for detection, as fraudsters themselves are using state-of-the-art technology to avoid detection such as artificial intelligence to change behavior in real-time and CLI Spoofing for more successful Robocall attacks.
By using Mavenir’s native ML algorithms to identify fraud and other anomalous network behavior, reliance on rules can be avoided, providing higher accuracy, lower false positives, and the knowledge that known, future, and unknown types of fraud will always be quickly identified. CallShield’s framework allows flexible control of all processing and decision stages via rules leveraging both ML and Rule-Engine technologies.
CallShield is delivered with three ML models to automatically detect and classify behavioral anomalies across voice fraud (IRSF, Wangiri), robocalling, and call centers.
Within CallShield, features used by ML include classifying abnormal traffic peaks, regular interval ranges, anomalous behavior classification, answer rates, voicemail redirect rates, typical duration patterns, social graph analysis, and other unknown anomalies. Dedicated detection of neighbor spoofing, mirror spoofing, and enterprise spoofing techniques are also supported.
ML identifies this traffic for detailed monitoring, and analyzes all calls, the caller-id initiates to callee’s sharing the same range. Deep LSTM and CNN networks review entire neighborhoods holistically to identify Robocall / Nuisance Calls, and quickly identify clusters of suspect calls with near-perfect accuracy, even if caller-ids are spoofed. This is completed through a combination of behavioral analysis and feature categorization.
Protection against CLI Spoofing is supported with prebuilt spoofing detection techniques.
CLIs are examined to identify inaccuracies that may indicate spoofing, such as CLI lengths that are too long or too short, or even of an incorrect format or from an unallocated number range or fixed area code.
CLI origination can also be examined to determine the validity of local CLI’s originating from an offnet interconnect by performing on-net presence checks.
Fraudulent Call Centers
CallShield’s ML supports dedicated features for fraudulent call centers, operating from dedicated bases and generating mass nuisance calls to subscribers. These call centers are often focused on specific frauds or scams.
ML features support dedicated detection of these call centers in operation, including outbound call rate, declining call rate, voicemail durations, and call origin.
Wangri, and Voice Fraud (IRSF)
CallShield features used with ML include identifying sudden increases in traffic.
Detection factors include typically uncommon destination, time of calls, history of communication from the number, duration, time between calls, connection length, roaming status, IMEI change, and average call duration.
The Distinct Advantage: CallShield
Mavenir has brought the latest in detection together from 10+ years of proprietary Machine Learning technology, enhanced with a fully scalable and highly adaptable solution suitable for CSPs, MNOs, MVNOs, and wholesalers alike.
- Real-time detection, analysis, blocking, and subscriber warning
- Continuous real-time learning with a dynamic Machine Learning model
- Enterprise support with Caller ID enrichment and management, with optional monetization
- Pre-integrated with Mavenir traffic-based solutions
- Diverse deployment options based on unique needs, including SaaS through Mavenir SMART Services
Three Rules of Success in Building a Cloud-Native Telco
30 September 2022
A cloud-native network opens several new possibilities for the service providers to deliver a differentiated experience and accelerate the launch of new services even as they bring down the total cost of ownership of running the network.
Mavenir to Deliver Cloud-Based 5G Solutions on Google Cloud
30 September 2022
Mavenir announces integration of cloud-native 5G solutions with public cloud infrastructure on Google Cloud, expanding on Mavenir’s “One Network, Any Cloud All Software” strategy.