The Internet of Things (IoT) ecosystem connects physical devices to the internet, offering significant advantages in agility, responsiveness, and potential environmental benefits. The number and variety of IoT devices are sharply increasing, and as they do, they generate significant data sources. Deep learning (DL) algorithms are increasingly integrated into IoT applications to learn and infer patterns and make intelligent decisions. However, current IoT paradigms rely on centralized storage and computing to operate the DL algorithms. This key central component can potentially cause issues in scalability, security threats, and privacy breaches. Federated learning (FL) has emerged as a new paradigm for DL algorithms to preserve data privacy. Although FL helps reduce privacy leakage by avoiding transferring client data, it still has many challenges related to models’ vulnerabilities and attacks. With the emergence of blockchain and smart contracts, the utilization of these technologies has the potential to safeguard FL across IoT ecosystems. This study aims to review blockchain-based FL methods for securing IoT systems holistically. It presents the current state of research in blockchain, how it can be applied to FL approaches, current IoT security issues, and responses to outline the need to use emerging approaches toward the security and privacy of IoT ecosystems. It also focuses on IoT data analytics from a security perspective and the open research questions. It also provides a thorough literature review of blockchain-based FL approaches for IoT applications. Finally, the challenges and risks associated with integrating blockchain and FL in IoT are discussed to be considered in future works.

Federated learning (FL) enables distributed joint training of machine learning (ML) models without the need to share local data. FL is, however, not immune to privacy threats such as model inversion (MI) attacks. The conventional FL paradigm often uses privacy-preserving techniques, and this could lead to a considerable loss in the model’s utility and consequently compromised by MI attackers. Seeking to address this limitation, this paper proposes a robust variational encoder-based personalised FL (RVE-PFL) approach that mitigates MI attacks, preserves model utility, and ensures data privacy. RVE-PFL comprises an innovative personalised variational encoder architecture and a trustworthy threat model-integrated FL method to autonomously preserve data privacy, and mitigate MI attacks. The proposed architecture seamlessly trains heterogeneous data at every client, while the proposed approach aggregates data at the server side and effectively discriminates against adversarial settings (i.e., MI); thus, achieving robustness and trustworthiness in real-time. RVE-PFL is evaluated on three benchmark datasets, namely: MNIST, Fashion-MNIST, and Cifar-10. The experimental results revealed that RVE-PFL achieves high accuracy level while preserving data and tuning adversarial settings. It outperforms Noising before Model Aggregation FL (NbAFL) with significant accuracy improvements of 8%, 20%, and 59% on MNIST, Fashion-MNIST, and Cifar-10, respectively. These findings reinforce the effectiveness of RVE-PFL in protecting against MI attacks while maintaining the model’s utility. The source code for RVE-PFL can be found on GitHub 1.

Federated learning (FL) has become a promising framework for collaborative model training on devices while preserving privacy. However, despite its significant potential, it faces notable cyber threats, such as poisoning attacks and codenamed Byzantine clients. These threats have the potential to significantly degrade the global model by jeopardizing the integrity of the collaborative model training process. Whilst previous research has addressed the detection and elimination of malicious gradients from Byzantine clients, it has also shown that model poisoning attacks can evade most statistical defence approaches relying on metrics such as median and distance. To address the challenge posed by poisoning attacks, we introduce a novel approach called Layerwise Gradient Purify (LGP), which aims to remove any harmful gradients before the global aggregation process. It is comprised of two closely-related stages. The first stage focuses on pruning gradients at the layer level using the Median Absolute Deviation (MAD) pruning criterion. In the second stage, statistical features are extracted from the pruned gradients layer-by-layer and then clustered into honest and malicious categories. The proposed methodology treats each layer of the model across all clients as a probability distribution, employing hierarchical clustering to differentiate between malicious and honest clusters. Moreover, we introduce a new innocent criterion for selecting honest clusters, relying on reputation scores and gradient deviations from the global model. Extensive experiments were conducted employing diverse deep learning models, including CNN, RNN, and MLP, across a spectrum of datasets, including Cifar-10, AG-News, MNIST and ToN-IoT. The experiments evaluated the resilience of state-of-the-art approaches against recently introduced attacks. The numerical results demonstrate that the LGP approach is effective and superior.

Cyberworthiness extends the concept of cybersecurity by evaluating whether systems and networks can perform their intended functions securely while maintaining protection against cyber threats. In corporate environments, cyberworthiness aims to ensure security, operational resilience, and trustworthiness across interconnected business processes and digital infrastructures. Modern organisations increasingly rely on complex cyber–physical and information systems, where vulnerabilities in software, networks, and devices can introduce significant operational and security risks. Cyberworthiness, therefore, encompasses security controls, risk management practices, and compliance with recognised cybersecurity standards and governance frameworks. It supports the assessment of information technology components and their exposure to both known and emerging cyber attacks, enabling organisations to evaluate system robustness and operational continuity. While cyberworthiness has historical foundations in system assurance and dependability, it also provides a conceptual basis for contemporary cyber resilience strategies. This paper discusses the concept of cyberworthiness in corporate organisations and identifies potential pathways for its practical implementation. It analyses existing cybersecurity standards and governance frameworks to support structured cyberworthiness assessment. This study presents a structured comparative review of fifteen cyberworthiness-relevant standards, supported by a Source Quality Appraisal Framework, a Framework Selection Guide specifying when each standard should be preferred and where conflicts arise, and a five-dimensional Cyberworthiness Assessment Readiness Model (CARM), a directional self-assessment instrument. The Efficient Automatic Safety and Security Assurance (EASSA) concept is proposed as a direction for future research, not a validated deployed system. Ensuring cyberworthiness remains challenging due to automation limitations in all reviewed standards, evolving threat landscapes, and governance complexity, requiring organisations to adopt integrated and measurable approaches to safeguard their digital assets and operational systems.