AWS Builder Center
PRISM: Predictive Real-time IoT Security Monitoring

PRISM: Predictive Real-time IoT Security Monitoring

A capstone project simulating IoT anomaly detection and incident response — and a look at how it could scale using AWS.

Cybersecurity & AiML
PRISM is a simulated IoT anomaly detection and incident response system built as a final-year capstone. The pipeline starts with an IoT simulation layer using Node-RED and Python scripts, streaming data over MQTT via Eclipse Mosquitto. On the detection side, Isolation Forest and Autoencoder models — trained on the TON_IoT and UNSW-NB15 datasets — flag anomalous device behavior in real time.
Detected anomalies are mapped against NIST CSF and MITRE ATT&CK for ICS to give each alert security context, not just a raw score. A Flask/React dashboard ties it together with live monitoring, built out across a 7-phase roadmap over 14 weeks.
Right now, PRISM runs as a self-contained simulation — everything from the broker to the ML models lives on local infrastructure. That's the right call for a capstone timeline, but it also maps cleanly onto a real AWS architecture if the project ever needed to scale beyond a lab environment: • AWS IoT Core could replace the local Mosquitto broker, handling device authentication and MQTT ingestion at scale. • Kinesis Data Streams could take over from the current pipeline for buffering and routing sensor data to the detection layer in real time. • SageMaker could host and retrain the Isolation Forest/Autoencoder models, replacing manual retraining with scheduled or triggered jobs as new attack data comes in. • Lambda fits naturally as the glue layer — triggering inference on incoming Kinesis records or routing confirmed anomalies to an alerting pipeline. • GuardDuty for IoT devices could complement PRISM's custom models with AWS's own managed threat intelligence, giving a second detection layer to cross-reference against. • DynamoDB or Timestream would handle the anomaly/event log at scale better than a local database, especially for the time-series nature of sensor data. • QuickSight could replace or extend the Flask/React dashboard for a managed, shareable monitoring view.
None of this is required to make PRISM work as a capstone — but it's a natural "what's next" if the project ever moves from a controlled lab simulation toward something deployable.
Any opinions in this article are those of the individual author and may not reflect the opinions of AWS.
Enjoyed reading this content? Let the author know!

Your likes, comments, shares, and saves help creators reach more builders.

Loading recommendations

Loading article