SUPR
Federated-learning-based anomaly detection for GPS
Dnr:

NAISS 2024/22-1232

Type:

NAISS Small Compute

Principal Investigator:

Wenjie Liu

Affiliation:

Kungliga Tekniska högskolan

Start Date:

2024-09-23

End Date:

2025-10-01

Primary Classification:

20206: Computer Systems

Webpage:

Allocation

Abstract

I will try a federated learning framework based on my previous GNSS attack detection scheme. To be more specific, in the previous scheme, I train a detector by using a collection of all GNSS locations with signals’ properties in a centralized machine. Now, I am going to train local models using local GNSS data and then aggregate the local models into a stronger global model. I will focus on the framework design and explore both independent and identically distributed (i.i.d.) and non-i.i.d. GNSS data across different mobile devices, where i.i.d. data means training data is from the same trace (distribution) as testing data and similarly, non-i.i.d. data means training data is from the different trace (distribution) as testing data.