Skip to main content
Version: v4.0.5 [Denim]

Lab 4: End-to-End Anomaly Detection with Interference Injection

This lab demonstrates a full end-to-end experiment combining:

  • 5G network deployment
  • External UE attachment (over-the-air)
  • Uplink traffic generation
  • MX-AI anomaly detection inference
  • Intentional interference injection using a second network
  • Natural-language explainability via agents

The goal is to reproduce an interference scenario and observe how the anomaly detection agent detects, explains, and reasons about the issue.

Experiment workflow​

Prerequisites​

Before starting this lab, make sure you have:

  • Access to a BubbleRAN cluster
  • SSH access to the cluster nodes
  • Two Quectel external UEs (e.g. Quectel RM520N-GL)
  • Two USRP B210
  • An Ubuntu machine to host the UE Snap

Step 1 — Deploy the First Network (Baseline Network)​

This network represents the normal operating condition. The anomaly detection model will implicitly treat this phase as the baseline behavior.

Network and UE Definition​

Create a YAML file (e.g. network_1.yaml) .

⚠️ Important:

  • Keep the band (n41) unchanged.
  • Set the annotation extras.t9s.io/scenario: 'network_1'.
  • Set the worker name in scheduling to specify where you deploy the network.
  • This value is used by the inference pipeline to attach to the correct scenario.
apiVersion: athena.trirematics.io/v1
kind: Network
metadata:
  name: bubbleran1
  namespace: trirematics
spec:
  slices:
    - plmn: "00101"
      dnn: "internet"
      network-mode: "IPv4"
      service-type: eMBB
      differentiator: 0x000000
      ipv4-range: "12.1.1.0/24"
      ipv6-range: "2001:db8:1::/64"
  access:
    - name: oai-gnb
      stack: 5g-sa
      model: oai-ran/monolithic-gnb
      identity:
        an-id: 50
      radio:
        device: uhd-b200
      scheduling:
        nodeName: <worker_node_name>
      cells:
        - band: n41
          arfcn: 504990
          bandwidth: 40MHz
          subcarrier-spacing: 30kHz
          tdd-config:
            period: 5ms
            dl-slots: 7
            dl-symbols: 6
            ul-slots: 2
            ul-symbols: 4
      controller: flexric.bubbleran1
      core-networks:
        - minimal.bubbleran1
  core:
    - name: minimal
      stack: 5g-sa
      model: open5gs/5gc
      identity:
        region: 0
        cn-group: 4
        cn-id: 5
  edge:
    - name: flexric
      stack: 5g-sa
      model: mosaic5g/flexric
    - name: monitoring
      stack: 5g-sa
      model: mosaic5g/monitoring-python
      profiles:
        - rlc-sm
        - pdcp-sm
        - mac-sm
        - gtp-sm
        - slice-sm
        - kpm-sm
        - database
      annotations:
        extras.t9s.io/scenario: 'network_1'
---
apiVersion: athena.trirematics.io/v1
kind: Terminal
metadata:
  name: ue1
  namespace: trirematics
spec:
  vendor: ogs
  stack: 5g-sa
  model: terminal/external
  target-cores:
    - minimal.bubbleran1
  identity:
    imsi: "001010000000111"  # replace with the IMSI corresponding to your SIM card
    pin: "1234"
    opc: "0xC42449363BBAD02B66D16BC975D77CC1"
    key: "0xfec86ba6eb707ed08905757b1bb44b8f"
    sqn: "0xff9bb4000001"
  slice:
    dnn: internet
    network-mode: IPv4
    service-type: eMBB
    differentiator: 0x000000
  radio:
    bands:
      - n41

Deploy it:

brc install network network_1.yaml
  • Check for the status of the deployment using the command brc observe. Wait until all the Elements are in the STATUS set to 1/1 Y state.

Step 2 — Attach the External UE (Over-the-Air)​

Because this is an external UE, it must be attached manually using the UE Snap.

On the Ubuntu Machine Connected to the Quectel UE​

sudo snap install --devmode new-oai-ue.snap
snap list | grep oai-ue
sudo oai-ue.init
sudo oai-ue.conf edit quectel

Edit the configuration:

  • Set PLMN = 00101

  • Band = n41

  • ARFCN = 504990

  • ⚠️ Make sure to use the same IMSI consistently in the UE configuration, matching the IMSI of the SIM card being used.

Grant USB access:

sudo snap connect oai-ue:raw-usb

Start the UE:

sudo oai-ue.quectel

From the machine where the UE is connected

oai-ue.iperf3 -c 12.1.1.1 -t 10000s
  • You can open a new terminal (command line) to generate traffic, while keeping the first terminal for deployment.

This traffic represents normal network behavior.

For more details on external UE over-the-air deployment, refer to the official BubbleRAN documentation:
https://bubbleran.com/docs/test-cases/over-the-air/ue/#external-deployment

Step 3 — Deploy the MX-AI Anomaly Detection Fabric​

This deploys:

  • UI
  • Orchestrator
  • Uplink anomaly detection agent
  • Anomaly database (MCP)

AIFabric Definition​

Create a YAML file (e.g. myfabric.yaml)

apiVersion: odin.trirematics.io/v1
kind: AIFabric
metadata:
  name: myfabric
  namespace: trirematics
spec:
  ui: hub.bubbleran.com/orama/ui/iris:latest
  mcp:
    - name: anomaly-db
      image: hub.bubbleran.com/orama/mcp/anomaly-db:latest
      env:
        - name: ANOMALY_SCENARIO
          value: network_1
  llms:
    - name: openai-model
      provider: openai
      model: gpt-4.1-mini
      apiKey: <YOUR_API_KEY>
  topology: custom
  agents:
    - name: supervisor
      role: worker
      image: hub.bubbleran.com/orama/agents/supervisor:latest
      llm: openai-model
      dependencies:
        - ul-ad-agent
      icp: a2a
    - name: ul-ad-agent
      role: worker
      image: hub.bubbleran.com/orama/agents/ul-ad-agent:latest
      llm: openai-model
      icp: a2a
      mcpServers:
        - anomaly-db

Deploy it:

brc install aifabric myfabric.yaml

Step 4 — Access the UI​

# If you are NOT already inside the cluster, first establish an SSH tunnel
ssh -L 9900:localhost:9900 <user>@<cluster-ip>

# Then, inside the cluster, forward the UI service port
kubectl -n trirematics port-forward service/myfabric-ui 9900:9900

Open:

http://localhost:9900

Example questions:

  • "Is there currently an anomaly in the network?"
  • "Is there an anomaly impacting users right now?"

At this stage, the system should report normal behavior.

Step 5 — Deploy the Second Network (Interference Source)​

This network intentionally shares the same band and ARFCN to generate RF interference.

⚠️ Important

  • Set the worker name in scheduling to specify where you deploy the network.
apiVersion: athena.trirematics.io/v1
kind: Network
metadata:
    name: bubbleran2
    namespace: trirematics
spec:
    slices:
        -   plmn: "00102"
            dnn: "internet"
            network-mode: "IPv4"
            service-type: eMBB
            differentiator: 0x000000
            ipv4-range: "12.1.1.0/24"
            ipv6-range: "2001:db8:1::/64"
    access:
        -   name: oai-gnb
            stack: 5g-sa
            model: oai-ran/monolithic-gnb
            identity:
                an-id: 10
            radio:
                device: uhd-b200
            scheduling:
                nodeName: <worker_node_name>
            cells:
                -   band: n41
                    arfcn: 504990
                    bandwidth: 40MHz
                    subcarrier-spacing: 30kHz
                    tdd-config:
                        period: 5ms
                        dl-slots: 7
                        dl-symbols: 6
                        ul-slots: 2
                        ul-symbols: 4
            core-networks:
                - minimal.bubbleran2
    core:
        -   name: minimal
            stack: 5g-sa
            model: open5gs/5gc
            identity:
                region: 0
                cn-group: 4
                cn-id: 5
    dns:
        ipv4:
            default: 8.8.8.8
            secondary: 8.8.4.4
---
apiVersion: athena.trirematics.io/v1
kind: Terminal
metadata:
    name: ue2
    namespace: trirematics
spec:
    vendor: ogs
    stack: 5g-sa
    model: terminal/external
    target-cores:
        - minimal.bubbleran2
    identity:
        imsi: "001020000002022" # replace with the IMSI corresponding to your SIM card
        pin: "1234"
        opc: "0xC42449363BBAD02B66D16BC975D77CC1"
        key: "0xfec86ba6eb707ed08905757b1bb44b8f"
        sqn: "0xff9bb4000001"
    slice:
        dnn: internet
        network-mode: IPv4
        service-type: eMBB
        differentiator: 0x000000
    radio:
        bands:
          - n41

Deploy it:

brc install network network_2.yaml
  • Check for the status of the deployment using the command brc observe. Wait until all the Elements are in the STATUS set to 1/1 Y state.

Deploy the network and external UE similarly, ensuring:

  • Band: n41

  • ARFCN: 504990

  • Set PLMN = 00102

  • ⚠️ Make sure to use the same IMSI consistently in the UE configuration, matching the IMSI of the SIM card being used.

Deploy traffic:

oai-ue.iperf3 -c 12.1.1.1 -t 1000s

This creates uplink interference impacting the first network.

Step 6 — Observe Detection & Explainability​

Return to the UI and ask:

  • "What happened in the last 5 minutes?"
  • "What is happening in the network right now?"
  • "What is the severity of the anomaly?"
  • "Which KPIs are deviating and what do they mean?"
  • "What is the most likely cause?"
  • "Given that cause, what are the 3 best next checks or actions?"

The agent should:

  • Detect the anomaly
  • Attribute it to uplink interference
  • Explain affected KPIs
  • Suggest concrete next actions

Outcome​

You have successfully:

  • Deployed a real 5G Over-the-air setup
  • Triggered controlled interference
  • Observed real-time anomaly detection
  • Interacted with the system using natural language
  • Validated explainability and decision support
Last updated on by Risma