Dependable Artificial Intelligence and Cyber Security

Progress in Artificial Intelligence (AI), Technologies such as Machine Learning (ML) and Generative AI, have has resulted in them being integrated with every application area. While the AI/ML techniques may be subject to cyber-attacks, the AI/ML algorithms are also being applied to solve cyber security problems. This book covers both AI/ML for cyber security as well as ensuring that AI/ML systems are secure, ensure privacy and integrity, among others. The AI/ML techniques that ensure security, integrity, privacy, fairness, and are fault tolerant and resilient have come to be known as Dependable Artificial Intelligence (DAI) and Dependable Machine learning (DML).

This book addresses various aspects of DAI/DML with cyber security applications as the main focus area. This book is divided into five parts, each describing some aspect of the technology that is relevant to DML and DAI. Part I describes supporting technologies for DML and DAI including security and privacy, an overview of various ML and AI techniques, applying AI/ML for Cyber Security , and (ii) Big Data, Cloud Computing, IoT and Blockchain technologies. Part II describes various concepts and techniques in DML. In particular, we discuss a framework for DML, different adversarial machine learning techniques, such as support vector machines and relevance vector machines, Privacy aware ML techniques such as privacy aware synthetic data generation and privacy aware federated learning with blockchain, Fairness in ML, and aspects such as integrity and resilience for ML applications.

Part III describes secure cloud-based data infrastructures for DML. We focus on Cloud-based infrastructures due to the extensive experimental work we have carried out in this field and many of the DML systems are currently hosted on cloud platforms. In particular, we discuss cloud-based infrastructure for secure cloud-based query processing infrastcrtures, assured information sharing infrastructures, data integration infrastructures to host DML models, systems and applications. Part IV discusses applications of DML in cyber security. The cyber security applications we have discussed include malware detection and insider threat detection. We also discuss how attackers could use ML to penetrate into various computing systems. Part V focusses on various DAI technologies and systems. In particular, we describe a framework for DAI, securing LLMs (Large Language Models), applying LLMs for security, as well as aspects of AI Governance.

The book is concluded in Chapter 43 with a discussion of future directions. It also includes two appendices, one which shows how we have evolved from data management to data mining to data science to ML to AI, and other providing readings in Dependable AI and Cyber Security.