DIALLO Abdoulaye

Affiliated project IoT-SDC

Affiliated CEAs :

CEA SMIA, University of Abomey-Calavi

CEA MITIC Gaston Berger University

Thesis supervisor: Prof. Chérif DIALLO, ACE MITIC, UFR des Sciences Appliquées et de Technologies, Université Gaston Berger, Senegal, cherif.diallo@ugb.edu.sn

Thesis co-supervision: Prof. Eugène C. EZIN, ACE SMIA, Institut de Mathématiques et de Sciences Physiques, Université d'Abomey-Calavi, Benin, eugene.ezin@gmail.com

Other contributors to thesis supervision: Prof. Mady CISSE, ESP, UCAD, Senegal

Start date: 01/06/2021
Anticipated date of thesis defense: June-July 2024
ORCID profile: 0000-0003-0591-4525

Title: Dynamic framework for preventing and responding to security incidents in the IoT

• LinkedIn

Summary of your doctoral topic:

The Internet of Things (IoT) has revolutionized our daily lives, streamlining tasks in a variety of areas. However, the proliferation of mobile applications and the expansion of telecommunications networks have led to a massive increase in data flows, posing the problem of how to manage and classify them. In the field of cybersecurity, IoT exposes many devices to the risk of becoming the target of attacks such as botnets. An intelligent classifier, based on data analysis, is becoming crucial for detecting threats and responding rapidly to incidents. This project aims to develop a classification model for IoT data using artificial neural networks. This classifier will distinguish between normal and malicious traffic, and identify common types of attack in IoT systems. The significant contribution lies in the creation of a Dynamic Framework capable of preventing and responding to IoT security incidents using artificial intelligence techniques, in particular deep learning.

Summary of results achieved:
Our research results focus on security in the Internet of Things (IoT), with four publications covering smart home activity monitoring, anomaly detection and intrusion in IoT networks. The first paper [1] focuses on the application of deep learning techniques to the recognition of human activity in IoT environments, with a particular emphasis on smart home monitoring. We evaluated the effectiveness of different deep learning models, including MLP, RNN and LSTM, in classifying the activities of residents in a home equipped with ambient sensors. The second study [2] compares two classifiers, binary and multi-class, for anomaly detection, showing that the multi-class classifier has a slight advantage on data with known anomalies, while the binary classifier is more effective on data containing unknown anomalies. The third publication [3] proposes an intrusion detection system for MQTT traffic, using an autoencoder to detect intrusions with high accuracy on both normal and attack data. Finally, the fourth research [4] presents a three-level intrusion detection system, with autoencoders to detect intrusions and an MLP to accurately classify the nature of attacks. Overall, this work highlights the effectiveness of machine learning and deep learning-based approaches for enhancing the security of IoT networks.