Stage de Fin D'étude: Anomaly Detection in Iot - Issy-les-Moulineaux, France - Institut Supérieur d'Electronique de Paris (ISEP)
il y a 3 semaines
Description
_
Internship:
State of the art solutions for Anomaly Detection in IoT Environments_
General presentation / Technological context
The world is moving fast toward an era of Web 3.0.
The Internet of Things (IoT), part of such an evolution, connects smart devices to digital services for the convenience of consumer's life.
According to [1], the number of IoT devices might reach up to 75 billion connected to the Internet by the end of 2025.
However, this rapid increase of IoT devices leads also to the inheritance of the security, privacy, and trust problems, already well-known in traditional networks, making the IoT devices even more vulnerable due to both device heterogeneity and communication protocols constraints.
Open web Application Security Project (OWSAP) [2] has published a list of IoT vulnerabilities, which can be summarized into three categories: Device related threats, Communication channel threats, and Application and software-related threats.
Moreover, since such IoT devices collect and produce a huge amount of data, which requires a great deal of pre-processing, to better exploit them for training ML/DL models for instance.
But pre-processing raises some issues such as data availability, data quality in terms of accuracy and presence of errors or anomalies.
In this context, an anomaly is defined as any deviation from a normal behaviour modelled by the collected data, which might result from an external attack (such as a Man-in-the-Middle, or a data/model poisoning), from a compromised device, from a flaw or an implementation bug or simply from an abnormal contextual situation (e.g., human intervention, device failure, natural disaster, etc.).
Therefore, to be able to detect anomalies and protect such IoT devices from cyberattacks, different techniques were proposed in the literature using different approaches going from the logic-based (knowledge bases and ontologies) ones to the statistical ones (ML/DL).
In this internship, we focus on the statistical approaches to analyse, implement, evaluate, and compare different state-of-the-art machine learning algorithms for anomaly detection in IoT environments for a given dataset and a given use case to be defined.
Methodology
The idea is to compare different types of machine learning algorithms which can be used for anomaly detection in different IoT environments.
There are three main branches of machine learning:
Supervised, Semi supervised, and Unsupervised learning.
The first task is to study machine learning algorithms and select the best possible algorithms for detecting anomalies in IoTs.
As the data collected from the IoT devices and sensors are in raw form, the second task will be pre-processed the data according to the chosen machine learning algorithms.
Some possible IoT sensor datasets that can be used in the project are present in [7], and [8].The main steps include:
1) the pre-processing of the data, 2) the implementation, 3) training, and 4) testing of each ML algorithm. In the end, the result will be analysed based on the usual performance metrics (accuracy, recall, etc.).
Specifications / Tasks
- Screening the state of the art for relevant works
- Studying different ML algorithms and identifying the suitable one(s) for anomaly detection in IoT environments.
- Preprocessing/cleaning of the provided IoT dataset
- Implementation and evaluation of the algorithms
- Final comparison and conclusions
- Perspective: proposition of an enhanced solution for anomaly detection in IoT networks (scientific publication in a conference)
Prerequisites
- M2 in computer science, mathematics or equivalent
- Deep knowledge on Machine Learning / Deep Learning
- Knowledge in IoT is a plus
- A good level of English
- Interest and motivation to conduct research
How to apply?
Other information
- Duration and salary:
- 6 months starting March,
- Localisation : ISEP, 10, rue de Vanves IssylesMoulineaux.
- This internship might lead to a PhD thesis.
Type d'emploi :
Stage
Durée du contrat : 5-6 mois
Salaire : 595,35€ par mois
Type d'emploi :
Stage
Durée du contrat : 6 mois
Salaire : 595,35€ par mois
Avantages:
- Horaires flexibles
Exigences linguistiques flexibles:
- Français non requis
Programmation:
- Du lundi au vendredi
- Horaires flexibles
Lieu du poste :
Un seul lieu de travail
Date de début prévue : 15/03/2023
