Ralph KARAM : "Automatic detection of business data anomalies with deep learning and application to the ADS-B protocol"

Abstract : The use of Machine Learning (ML) and Deep Learning (DL) for security anomaly detection is an extremely active topic. Anomaly detection touches many domains, namely air traffic control, IoT, etc.
One main air traffic control technology is the ADS-B protocol (Automatic Dependent Surveillance-Broadcast). It constitutes an air traffic control data source based on satellite positioning. Each aircraft periodically sends via ADS-B messages its information to ground stations and other aircraft. ADS-B is becoming globally mandatory but it still lacks security measures like encryption and authentication. One way to tackle this problem is ML and DL based ADS-B anomaly detection. Supervised ADS-B anomaly detection was the main focus of this thesis due to its performance advantage compared to unsupervised methods. However, it suffers from the lack of labeled data. In order to obtain enough labeled anomalies and normal data, a false data generator was used. To the best of our knowledge, this thesis is the only work which used supervised ADS-B anomaly detection. Our approach gave very promising results in detecting various types of attacks. The best performance was obtained using the Long Short-Term Memory (LSTM) model. In addition, as a secondary contribution in this thesis, different DL and ML approaches were studied in order to forecast noise levels and detect punctual noise level anomalies. The data were gathered from an IoT system more specifically a network of smart parkmeters. The results of our study inferred that such methods, preferably a 1D Convolutional Long Short-Term Memory (CNN-LSTM), can be successfully used in environmental noise monitoring applications

Jury Composition :

Michel SALOMON, Maître de conférences, Université Bourgogne Franche-Comté, PhD Director

Sylvain CONTASSOT-VIVIER, Professeur des universités, Université de Lorraine, Reviewer

Raphaël COUTURIER, Professeur des universités, Université Bourgogne Franche-Comté, Reviewer

Stéphane CHRÉTIEN, Professeur des universités, Université Lumière Lyon 2, Reporter

Flavien VERNIER, Maître de conférences, Université Savoie Mont Blanc, Reporter

Localization : IUT Belfort-Montbéliard Site Belfort Techn’hom 19 avenue du Maréchal Juin BP 527 90016 BELFORT Cedex -  salle 108 du bâtiment E - Informatique