Back to portfolio
CybersecurityMachine-learning voice protection

PBX Threat Detection System

An AI-powered detection system that identifies and blocks toll fraud, unauthorized access, and suspicious PBX call patterns in real time.

99.7%
Detection Rate
<0.1%
False Positives
50ms
Avg Response

The challenge

What needed to change

Static PBX rules missed evolving fraud patterns and generated noisy alerts. Security teams needed fast detection that could adapt to normal calling behavior for each customer environment.

The solution

How we approached it

We built a streaming service that learns calling baselines, scores anomalous events, and applies automated controls through Asterisk and FreePBX integrations. Redis supports low-latency decisions and fast feature access.

Delivery highlights

What we delivered

Adaptive baselines for users, trunks, destinations, and time windows
Real-time scoring integrated with PBX call controls
Investigation views with explainable detection signals
Feedback workflows for continuous model refinement

Services

Threat researchMachine learningSecurity engineering

Business impact

Outcomes that lasted beyond launch

The implementation focused on measurable improvement and an operating model the client could continue to evolve.

01

Blocked high-risk calls before significant charges accumulated

02

Reduced alert fatigue through environment-specific models

03

Improved coverage without replacing existing PBX systems