In recent years, machine learning has set higher expectations from artificial intelligence (AI) technology, particularly, deep learning has shown exemplary performance in the field of Image recognition, natural language processing, pattern matching, face recognition, and many more. Deep Learning (DL) models has several benefits like fast computation of complex problems, maximum application of unstructured data, reduced costs, and many more but it has some limitations also associated with like opaqueness, computationally intensive, etc. However, the applications based on DL models are used in day-to-day routine and they work on huge amount of data to achieve higher accuracy and if these models lead to inaccuracies because of malicious activities, then it would become cumbersome and thus to protect the data from the security breaches is a major concern.
Deep learning (DL) models usually have sensitive information of the users and these models should not be vulnerable and expose to security and privacy. However, DL models are still susceptible to various security attacks perturbed by imperceptible noise which allow these models to forecast/ predict inaccurately with high degree of confidence. Therefore, it is important to look into the security aspects and related counter measure techniques of DL models. The book focuses on the recent advances and challenges related to the concerns of security and privacy issues in deep learning with an emphasis on the current state-of-art methods, methodologies and implementation, attacks, and their countermeasures. The book also discusses the challenges that need to be addressed for implementing DL-based security mechanisms that should have the capability in collecting or distributing data across several applications. The proposed volume will provide the deeper insights on deep learning models and security mechanisms across several applications. This book will unveil different applications of Metaheuristic approaches (i.e., swarm intelligence, genetic algorithm) in collaboration with DL models for high degree of confidence.