: CONTRIBUTIONS TO BIG DATAProcessing, Management, Security, and Exploitation of Data in Distributed and Dynamic Environments
Nadia Kabachi · HAL (Le Centre pour la Communication Scientifique Directe) · 2023
In recent years, we have witnessed a major revolution : Big Data, which has introduced a new paradigm in which science resides within the data. Research on Big Data covers a wide range of areas, including data processing and management, analysis, security, visualization, as well as legal and ethical aspects. This Habilitation thesis presents a non-exhaustive synthesis of my contributions in this field. The research I have conducted and supervised over the past decade has focused on four major axes:-The first one is related to the processing and curation of massive data. In this context, we have proposed an adaptive curation approach, based on services, to manage multi-source, multi-structured, and heterogeneous data collected in batch and streaming modes. This work has been materialized by the design and implementation of a service library for data extraction, enrichment, and quality evaluation.-The second axis concerns the management of massive data in the Business Intelligence context. Our interest has focused on the problems related to partitioning and distributing large volumes of data warehouses in a cluster of nodes according to the principles of the MapReduce programming model. To address these issues, we have proposed new schemes and strategies for data fragmentation and placement in a distributed and dynamic system such as Cloud Computing. Therefore, we have also enhanced the performance of the Hadoop framework.-The third axis of our research focuses on the security of massive data stored in the Cloud. The objective is to secure the storage of data by intentionally making them non-integral. In this perspective, we have developed an encryption algorithm based on data alteration. Intelligent multi-profile and adaptive agents to contexts and environment have also been designed to handle data alteration before storage and their « de-alteration » during queries for exploitation and analysis purposes.-The fourth axis of research focuses on the exploitation of data for recommendations, alerts, or decision-making purposes. We have made contributions through two approaches. The first one concerns the development of an alert system based on agents that aims to process heterogeneous and multi-source data. Its objective is to detect unusual facts or weak signals and alert accordingly. The second one focuses on the proposal of an explained recommendation model for crisis management. This model is based on deep learning, which exploits both different data sources and usage context to recommend management measures suitable for each situation. Additionally, it uses a semantic representation that provides adaptive explanations (XAI) of recommendations based on the context, role, and preferences of users. We have validated these contributions in the health field by applying them to the detection of emerging diseases and the recommendation of health measures.All our solutions have been developed and tested on massive, synthetic, and real data, mainly in the field of healthcare.Finally, this thesis highlights the priority research directions that I plan to explore. Among them, priority is given to the dynamic and intelligent protection of data 'throughout their lifecycle', ensuring their confidentiality, integrity, and availability. Furthermore, we also focus on the processing and exploitation of large volumes of real-time data for predictive maintenance, relying on continuous learning and Neuroevolution approaches.