Akilapa Guardian

AI Technology

Security Anomaly Detection

AI-powered anomaly detection across GPS telemetry, premises sensors, and access control data — identifying statistically unusual patterns that may indicate security incidents before they escalate.

94%True positive rate on anomaly detection (2024 audit)

How it works

Anomaly Detection establishes a behavioural baseline for each asset and location, then continuously monitors for statistically significant deviations from that baseline.

Step 1

Baseline Learning

The system observes 21 days of normal operational patterns — typical routes, stop durations, access times, and movement speeds — to establish a behavioural baseline per asset.

Step 2

Continuous Deviation Analysis

Real-time data is compared against the baseline using a statistical anomaly model. Deviations are scored by magnitude — small deviations ignored, large deviations escalated.

Step 3

Context-Aware Alerting

The system uses contextual intelligence to reduce false positives — for example, a vehicle stopping at a known fuel station is not flagged, but stopping at the same location at 02:00 is.

Data inputs

  • GPS telemetry and stop patterns
  • Access control event logs
  • CCTV motion event data
  • Historical behavioural baselines
  • Operator-defined normal hours and zones

Outputs

Anomaly Alert

Real-time notification for statistically significant deviations

Anomaly Score

Severity score (1–10) for each detected anomaly

Evidence Package

Timestamped evidence bundle including GPS, video clip, and access log for each anomaly

Trend Report

Weekly summary of anomaly patterns for security review

Example insight

Anomaly detected: Vehicle LGS-BUS-007 stopped at an unregistered location on Lekki-Epe Expressway for 47 minutes at 23:12 — significantly outside baseline (average stop duration: 4 minutes). SOC reviewing.

See Security Anomaly Detection on your fleet

Talk to our security architects about vehicles, fleets, churches, schools, and enterprise assets.