Federated Artificial Intelligence (FAI) is being rapidly adopted in life sciences, namely with drugs and genomics, and enables institutions to develop machine-learning models while preserving confidentiality around sensitive patient information. This chapter provides an overview of FAI's role in speeding up large-scale drug discovery and genomic research through privacy-preserving collaborative learning. A federated approach has been developed that combines the genomics and bioinformatics resources of several pharmaceutical labs and genomics repositories for the purpose of biomarker discovery, compound screening, and personalised treatment design. The FAI framework can achieve an average of over 91% predictive accuracy and an average of approximately 85% less overhead in terms of data transfer as compared to traditional centralised learning methods, all while maintaining confidentiality of all data. To validate the concepts presented in this work, real world examples of multi-institutional drug studies and genomic databases have been provided.
Due to the increasing complexity of cyberattacks, old-school cybercrime detection methods are no longer effective. This chapter looks at how machine learning-based intrusion detection systems and anomaly detection solutions can revolutionise the way cybersecurity is done today. By using various different machine learning (ML) algorithms, such as supervised, unsupervised, and reinforcement learning and find outliers (e.g. identify anomalies), detect new types of threats, and automatically respond to security incidents. This chapter provides a thorough overview of how ML is used for intrusion detection, including the use of deep learning algorithms for threat intelligence, behavioural and static analysis of malware, identifying phishing schemes using natural language processing, and detecting zero-day attacks. Additionally, this chapter dealing with unbalanced datasets, defending against adversarial attacks, using federated learning to cover multiple locations, and enabling explainability for ML-based cybersecurity.