Logo Lanfrica

CharlesKariuki-001/africandr

Domaine:

digital infrastructure

Type de record:

software
Créateur:
Cha
Hôte:
AI-powered Network Detection & Response for African mobile money infrastructure. Zeek + ML + anomaly detection. Part of Vigilant AI. # 🛡️ AfricaNDR AI-Powered Network Detection & Response Built for African infrastructure. Powered by AI. Deployable anywhere. Network level threat detection, built for African fintech infrastructure. ## The problem Most fraud detection only looks at transactions after they happen. By the time a fraud detector reads a message or a transfer, the attack that made it possible has usually already succeeded. A SIM swap, for example, happens at the network and telecom level, hours or days before a single fraudulent transaction is ever attempted. If nothing is watching that layer, the fraud detector downstream never had a chance. SACCOs, small fintechs, and mobile money agents in Kenya mostly run on infrastructure with no dedicated network monitoring at all. Big banks can afford enterprise network security tools. Everyone else operates blind. AfricaNDR is built to give that same visibility to the institutions that currently have none. ## What it watches for **General network threats** Port scanning, command and control beaconing, DNS tunneling, lateral movement inside a network, data leaving where it should not. **Threats specific to African mobile money** SIM swap attempts visible at the network level before they complete, automated abuse of mobile money APIs, unusual USSD session behavior that signals an account takeover in progress, compromised agent accounts, transaction velocity that looks wrong at the network layer, before it even reaches a transaction record. ## How it works Traffic gets captured using Zeek, a network monitoring tool, which turns raw traffic into structured logs. Those logs get run through custom feature extraction, then an anomaly detection model flags what looks wrong. Anything flagged shows up on a live dashboard, and the same signal gets passed to Vigilant AI, where it combines with message level fraud scoring. ``` Network traffic down to Zeek sensor, turns raw traffic into structured logs down to Feature extraction, flow level and pat …