Secured AI-Based Multimedia Communication
Mahmoud A. Shawky, Divyanshu Awasthi, Ahmed Adel Ramadan, Sameh Zarif, Jawad Elsayed Ahmad, Syed Tariq Shah · 2025
The fast-growing use of artificial intelligence (AI) in multimedia communication systems has generated seismic change across areas such as smart cities, e-health, and environmental monitoring. While the advantages are clear, and efficiencies generated through AI integration are significant, they can also create severe vulnerabilities—in this case, adversarial attacks and deepfakes, as well as energy inefficiencies that hinder security and global sustainability efforts. This chapter provides a framework for secured AI-based multimedia communication that links advancement with sustainable development and security measures. By integrating AI technologies with cybersecurity methods and environmental impact assessment, it aims to achieve meaningful guidance for building resilient and energy-efficient systems. The chapter begins with a section underpinned by more fundamental definitions and general issues that surround AI-enabled multimedia processing technologies (e.g., convolutional neural networks (CNNs), transformers, generative adversarial networks (GANs)), as well as the emergence of 5G/6G and edge computing architectures in enabling high-fidelity communication with low latency. Then, it examines aspects of the rapidly changing landscape of security threats (e.g., data poisoning attacks, adversarial perturbations in visual and audio processing models, and breaches of privacy in biometric and internet-of-things (IoT)-enabled data environments) and supplies some potential defensive mechanisms, such as homomorphic encryption for privacy-preserving AI, blockchain-based decentralized trust approaches, and federated learning (FL)-based solutions that limit data exposure but maintain compliance with general data protection regulation (GDPR) and the AI Act. We then provide examples that illustrate the practical implementation of these approaches in smart cities (AI-based anomaly detection with blockchain), e-health (encrypted FL with diagnostic imaging), and climate (low-power analysis of remote satellite imagery). Next, a framework for assessing aspects of performance, such as robustness, explainable AI (XAI), cost of computation, and energy consumption, is proposed. This framework aligns with the United Nations (UN) sustainable development goals (SDGs). Finally, the chapter concludes with a research roadmap for secured and sustainable multimedia systems via quantum-safe AI, XAI, and cross-modal security for hybrid data streams.