Deep Learning-Based Speech Recognition System using Blockchain for Biometric Access Control.

Deep learning-based automatic speech recognition is a promising technology that has witnessed high interest from academics and industrials and proved its efficiency in several fields, including biometric access control systems. In such systems, the use of biometric data is challenging, namely in the centralized environment. It represents several risks, including information leakage, unreliability, security, privacy, etc. To tackle these issues, several approaches are proposed based on blockchain technology, which has recently gained popularity in various areas. Given the risks related to data security and many others to the technologies involved, it is essential today to have specific protection, particularly for authentication, access control, authorization, secure transfer, traceability, etc. To this end, this project aims to provide secure and intrusion-tolerant intelligent infrastructure and platform, combining the aforementioned technologies to assist and facilitate the daily and professional lives of human beings, respecting their privacy and guaranteeing the security of data and transactions. This research project aims more specifically at cybersecurity based on deep learning through the speech recognition system on the one hand. On the other hand, the ultimate goal is to detect anomalies in the behavior of users once accessing the system. The blockchain will be used to ensure the authentication and identification processes, to deploy the security policy with the smart contract, and also to take care of the generation of the keys necessary for the encryption. The authentication and identification process is done through a combination of voice recognition and user credentials (attributes), using deep learning to extract features and then comparing the template generated with those stored in the blockchain. In addition, our objective is to optimize the voice recognition algorithms. The optimization of such a system requires performance optimization of the platform (i.e., maximization) and risk reduction (i.e., minimization).