ADS-B Data Anomaly Detection with Machine Learning Methods

ADS-B data collected from a local antenna at TET (Thales Emarat Technologies), Abu Dhabi area, UAE, is used to propose anomaly detection models based on supervised (SML) and unsupervised machine learning (USML) algorithms. The data is collected, filtered, and extracted then used in generating two ty...

Full description

Saved in:
Bibliographic Details
Main Authors: Amine Azz, Mohammed El, Aljasmi, Abdalla, Fallah Seghrouchni, Amal El, Benzarti, Walid, Chopin, Philippe, Barbaresco, Frederic, Zitar, Raed Abu
Format: Conference Proceeding
Language:English
Subjects:
ADS-B
Anomaly Detection
Data models
Detectors
GRU
LSTM
Noise
Predictive models
Radio frequency
TCN
Training data
Vectors
Online Access:Request full text
Tags: Add Tag
No Tags, Be the first to tag this record!
Description
Summary:ADS-B data collected from a local antenna at TET (Thales Emarat Technologies), Abu Dhabi area, UAE, is used to propose anomaly detection models based on supervised (SML) and unsupervised machine learning (USML) algorithms. The data is collected, filtered, and extracted then used in generating two types of data; one is normal and the other one is synthesized as abnormal data. The Random Forest (RF), Support Vector Classification (SVC), and Logarithmic regression (LR) are used in the SML approach, while the Temporal Convolutional Networks (TCN), Long Term Short Term Memory (LSTM), and Gated Recurrent Unit (GRU) are used in the USLM approach. A prediction model is built in the USLM method with a heuristic as a scoring mechanism for anomaly detection. Simulations were conducted on both approaches and comparisons showed that the USML based on the TCN model with Zero Scoring detection heuristic obtained the best results.
ISSN:2575-7490
DOI:10.1109/MetroAeroSpace61015.2024.10591547