Page 373 - AI for Good Innovate for Impact

AI for Good Innovate for Impact

Use Case 17: Anomaly Detection Across 5G Slices Using Federated

Learning 4.3 - 5G

Country: India
Organization: Indian Institute of Technology, Hyderabad(IITH)

Contact Persons:
Dr. Antony Franklin (antony.franklin@ cse .iith .ac .in)
Mr. Yaswanth Kumar L S (cs24resch11007@ iith .ac .in)
Mr. Arjit Gupta (cs23mtech12001@ iith .ac .in)
Ms. Peddi Manognya (cs24mtech12020@ iith .ac .in)

1 Use Case Summary Table

Item Details
Category 5G

Problem to be Detecting anomalies in one 5G network slice and preventing their occur-
Addressed rence in other slices by sharing learned model parameters through
federated learning.

Key Aspects of the A decentralized anomaly detection system leveraging federated learn-
Solution ing to ensure knowledge transfer across slices while maintaining data
privacy.

Technology Keywords 5G, anomaly detection, federated learning, network slicing, AI-driven
security

Data Availability Self Generated

Metadata (Type of Network logs, slice performance metrics, anomaly detection signals
Data)

Model Training and Federated learning framework deployed across slices, utilizing anomaly
Fine-Tuning detection models trained on distributed slice data.

Testbeds or Pilot Open Source tools such as RAN+UE emulator and Free5GC
Deployments

2 Use Case Description

2�1 Description

Modern 5G networks use network slicing to allocate dedicated resources for different services.
However, anomalies—such as security breaches, misconfigurations, or unexpected traffic

337