A new unsupervised learning-based process for extraction of knowledge's and improving anomalies detection

Abstract

Lately, the world confronts day by day the birth of typically big data in various fields such as health, telecom, electronic sales, communication, service provision, industry, e-learning and e-commerce. This emergence was a brick and a major reason behind the realization of immense changes in the processes of manipulation and processing of data as the system of storage, processing, analysis and visualization of data. From an application point of view, the manipulation of such data requires the implication of certain numbers of usage standards such as those related to security, credibility and optimization of data. However, it seems so difficult to ensure the security of massive and gigantic data coming from a variety of sources, and especially in real time, because there is an almost total absence of methodologies ensuring the processing and protection of a volume of data automatically. In fact, our work takes stock of a new process allowing the secure handling of big data. This is about getting more out of machine learning, security patterns and decision trees. More specifically, our approach aims to design machines able of securing large data in real time while making them learn from a mapping of security patterns (learning data). This training will allow machines to develop in terms of security, which will allow them to improve their intelligence related to the detection of anomalies caused in the data during the data exploitation operation.