5G Security Testing for Autonomous Mining Vehicles

COMP6016 — Computer Science Project 2 | Curtin University | Semester 1, 2026

A full-stack private 5G security testbed for autonomous mining vehicles. The project simulates an Open5GS core, UERANSIM radio access, a physics-based telemetry simulator for Komatsu-style haul trucks, a MEC security API with tamper detection and ML-based scoring, and a Prometheus + Grafana observability layer.

---

Team — Group 1

Name	Student ID	Role
Hirdya Sharma	21749180	Infrastructure integration, physics simulator, MEC Edge API, telemetry pipeline
Deepika Sharma	21952195	Grafana dashboards, Prometheus observability
Achani Bandara	21741102	ML model training and model artifacts
Hitesh Pankhania	22471264	Attack execution and packet capture
Gurleen Kaur	22131597	Attack simulation and Wireshark analysis

Supervisors: Dr. Nasim Ferdosian · Dr. Reza Ryan

---

What’s in this version

This repository reflects the current working single-stack layout from the latest mining5g_setup zip.

Current live architecture

• Open5GS 5G core with Dockerised NFs
• UERANSIM gNB + 2 UE instances
• Physics simulator running telemetry/physics_simulator_real.py
• MEC API running mec/mec_edge_api.py
• Achani’s ML artifacts integrated into mec/
• Prometheus + Alertmanager + Node Exporter + Grafana
• Single docker-compose.yml orchestration

Important current runtime ports

These are the ports exposed by the current compose file:

Service	Container Port	Host Port
MEC Edge API	5000	5001
Physics Simulator	8000	8000
Prometheus	9090	9190
Node Exporter	9100	9110
Alertmanager	9093	9093
Grafana	3000	3000
AMF SCTP	38412/sctp	38413/sctp

Use http://localhost:5001/health for MEC health checks. The host port is 5001, not 5000.

---

Repository structure

mining5g_setup/
├── config/                      # Open5GS + UERANSIM configs
├── mec/                         # MEC API, model config, pickles, datasets
├── telemetry/                   # Physics simulator and telemetry exporter
├── monitoring/                  # Prometheus, Alertmanager, Grafana provisioning
├── mongo-init/                  # Open5GS subscriber bootstrap
├── docs/                        # Architecture diagrams and supporting assets
├── docker-compose.yml           # Full system orchestration
├── setup.sh                     # Fresh Ubuntu setup helper
├── test_everything.sh           # End-to-end validation script
├── fix_subscribers.sh           # Reinsert subscribers if Mongo volume is wiped
├── full_stack_test_report.txt   # Validation notes
└── README.md                    # This file

---

System architecture

Physics Simulator (10.0.0.40) ──POST /telemetry──► MEC Edge API (10.0.0.45)
            │                                           │
            └─────────────── /metrics ─────────────────►│
                                                        ▼
                                                Prometheus (10.0.0.50)
                                                        ▼
                                                   Grafana (10.0.0.60)

UERANSIM gNB/UEs  ───────────────► Open5GS Core ───────────────► MEC/Monitoring

Internal network map

Container	IP
mongo	10.0.0.2
open5gs-nrf	10.0.0.10
open5gs-ausf	10.0.0.11
open5gs-udm	10.0.0.12
open5gs-udr	10.0.0.13
open5gs-pcf	10.0.0.14
open5gs-nssf	10.0.0.15
open5gs-bsf	10.0.0.16
open5gs-amf	10.0.0.20
open5gs-smf	10.0.0.21
open5gs-upf	10.0.0.22
ueransim-gnb	10.0.0.30
ueransim-ue1	10.0.0.31
ueransim-ue2	10.0.0.32
telemetry-simulator	10.0.0.40
mec-edge	10.0.0.45
prometheus	10.0.0.50
node-exporter	10.0.0.51
alertmanager	10.0.0.53
grafana	10.0.0.60

---

Quick start

Option A — use the provided setup script

cd ~/Downloads
unzip mining5g_setup.zip
cd mining5g_setup
chmod +x setup.sh
sudo bash setup.sh

Option B — run directly if Docker is already installed

cd mining5g_setup
docker compose up -d --build

Check containers

docker compose ps

Key verification commands

# MEC health
curl http://localhost:5001/health

# Prometheus targets
curl http://localhost:9190/api/v1/targets

# Grafana
xdg-open http://localhost:3000

Grafana login:

• Username: admin
• Password: mining5g@secure

---

Open5GS and UERANSIM

Core components

The system uses gradiant/open5gs:2.7.0 containers for NRF, AUSF, UDM, UDR, PCF, NSSF, BSF, AMF, SMF, and UPF.

UEs

• ueransim-ue1 → IMSI 999700000000001
• ueransim-ue2 → IMSI 999700000000002

Quick checks

# UE registration

docker logs ueransim-ue1 --tail=20
docker logs ueransim-ue2 --tail=20

# UE tunnel interface

docker exec ueransim-ue1 ip addr
docker exec ueransim-ue2 ip addr

If Mongo volumes are reset and UEs stop registering, run:

bash fix_subscribers.sh

---

Physics simulator

Active simulator

The current stack runs:

python physics_simulator_real.py

inside the telemetry-simulator container.

Important note

• telemetry/physics_simulator_real.py = current simulator in use
• telemetry/simulator.py = older/basic simulator, not the main one for the current stack

What it does

The physics simulator models three autonomous mining vehicles and generates:

• speed
• fuel
• battery
• engine temperature
• payload
• RSRP
• SINR
• RTT
• packet loss
• uplink/downlink throughput
• GPS position

Telemetry is then posted to the MEC API at /telemetry and exported via /metrics.

Simulator checks

# container state

docker ps -a | grep telemetry-simulator

# startup command

docker inspect telemetry-simulator --format='{{.Config.Cmd}} {{.Config.Entrypoint}}'

# logs

docker logs --tail=50 telemetry-simulator

---

MEC Edge API

Current service location

• Container IP: 10.0.0.45:5000
• Host access: http://localhost:5001

Main endpoints

Method	Path	Purpose
GET	/health	Liveness, vehicles seen, model status
GET	/metrics	Prometheus scrape endpoint
GET	/events	Recent forensic/security events
POST	/telemetry	Ingest and validate telemetry
POST	/model/upload	Model hot-swap support

Current model files

Achani’s ML artifacts are stored directly in mec/:

• mec/model_attack.pkl
• mec/model_protocol.pkl
• mec/model_quality.pkl
• mec/scaler.pkl
• mec/mec_config.json

Health check

curl http://localhost:5001/health

Metrics check

curl http://localhost:5001/metrics | grep mec_

Valid telemetry test

curl -X POST http://localhost:5001/telemetry \
  -H "Content-Type: application/json" \
  -d '{
    "vehicle_id":"MV-TEST-1",
    "speed_kmh":22,
    "fuel_liters":2400,
    "battery_percent":84,
    "engine_temp_c":86,
    "rsrp_dbm":-72,
    "rtt_ms":22,
    "packet_loss_percent":0.8,
    "payload_tons":70,
    "sinr_db":21,
    "ul_mbps":18,
    "dl_mbps":75
  }'

Tamper / rejected telemetry test

curl -X POST http://localhost:5001/telemetry \
  -H "Content-Type: application/json" \
  -d '{
    "vehicle_id":"MV-TEST-2",
    "speed_kmh":56,
    "fuel_liters":2400,
    "battery_percent":84,
    "engine_temp_c":86,
    "rsrp_dbm":-72,
    "rtt_ms":22,
    "packet_loss_percent":0.8,
    "payload_tons":70,
    "sinr_db":21,
    "ul_mbps":18,
    "dl_mbps":75
  }'

The second payload should be rejected because the current tamper bounds treat speeds above 55 km/h as invalid.

---

ML integration

The current stack uses Achani’s model artifacts within the MEC service rather than a separate ML microservice.

Current integrated model capabilities

• attack prediction
• protocol prediction
• traffic quality prediction
• anomaly flagging
• tamper-aware rejection path

Useful metrics to inspect

curl http://localhost:5001/metrics | grep mec_

Typical metrics to expose or visualise include:

• mec_attack_prediction
• mec_attack_predictions_total
• mec_protocol_prediction
• mec_traffic_quality
• mec_achani_flag
• mec_anomaly_score
• mec_tamper_detected_total

---

Grafana and Prometheus

Prometheus

• URL: http://localhost:9190
• scrape interval: 5 seconds

Grafana

• URL: http://localhost:3000
• default dashboard file: monitoring/grafana/dashboards/mining-telemetry.json

Provisioning

Grafana is provisioned from:

• monitoring/grafana/provisioning/
• monitoring/grafana/dashboards/

Recommended security / AI panels

These are the most useful panels for the current stack:

• Attack Type Detections
• Predicted Attack per Vehicle
• Traffic Quality per Vehicle
• Protocol per Vehicle
• Achani Flag per Vehicle
• Anomaly Score per Vehicle
• Tamper Detections (5 min)
• Rejected Telemetry % (5 min)

---

Troubleshooting

MEC health fails on port 5000

Use port 5001 on the host:

curl http://localhost:5001/health

Telemetry simulator is stopped

docker ps -a | grep telemetry-simulator
docker logs --tail=100 telemetry-simulator
docker compose up -d --build telemetry-simulator

Only /metrics appears in MEC logs

Prometheus is scraping, but fresh telemetry may not be arriving. Check:

docker logs --tail=50 telemetry-simulator
docker logs -f mec-edge

You should see POST /telemetry entries when the simulator is active.

UE registration issues after wiping volumes

bash fix_subscribers.sh
docker compose restart ueransim-ue1 ueransim-ue2

Validate compose file

docker compose config >/dev/null

---

End-to-end test sequence

# 1. Start everything

docker compose up -d --build

# 2. Check services

docker compose ps

# 3. Check MEC API

curl http://localhost:5001/health
curl http://localhost:5001/metrics | grep mec_

# 4. Check Prometheus

curl http://localhost:9190/api/v1/targets

# 5. Open Grafana
# http://localhost:3000

---

Known practical notes

• The host MEC port is 5001, while the container listens on 5000.
• The physics simulator in active use is physics_simulator_real.py.
• Some simulator outputs can occasionally trigger tamper rejection if generated values exceed configured bounds.
• ML pickle files are already included in mec/ in this zip.
• This layout supersedes the older repo design that used separate mec-edge-api/, telemetrics/, prometheus/, and grafana/ folders.

---

References

• Open5GS documentation
• UERANSIM documentation
• Prometheus documentation
• Grafana documentation
• Komatsu 730E specification data
• 3GPP NR radio propagation references
• MITRE ATT&CK for ICS