Underwater drones are vital for scientific research, environmental monitoring, and maritime operations, allowing data collection in challenging environments. However, their deployment faces issues such as low bandwidth, high latency, signal attenuation, and intermittent connectivity due to mobility and water currents. Traditional centralized data processing approaches are inefficient under these conditions as they require transmitting large volumes of raw data to a central location. To address these challenges, this study proposes a Federated Learning (FL) framework specifically tailored for underwater networks. Unlike centralized approaches, FL enables underwater drones to collaboratively train a global intrusion detection model by processing data locally and sharing only model updates with the central server. This approach significantly improves data security by ensuring that sensitive information never leaves the local device, reducing the risk of interception or compromise during transmission. Furthermore, FL’s decentralized architecture inherently aligns with the dynamic and distributed nature of underwater drone networks. The proposed framework improves cyber intrusion detection by leveraging localized insights from individual drones to detect threats, including zero-day attacks, without directly exposing sensitive data. By preserving privacy and enabling collaborative anomaly detection, FL addresses key cybersecurity challenges in the Internet of Underwater Things (IoUT).

Telehealth systems have evolved into more prevalent services that can serve people in remote locations and at their homes via smart devices and 5G systems. Protecting the privacy and security of users is crucial in such online systems. Although there are many protocols to provide security through strong authentication systems, sophisticated IoT attacks are becoming more prevalent. Using machine learning to handle biometric information or physical layer features is key to addressing authentication problems for human and IoT devices, respectively. This tutorial discusses machine learning applications to propose robust authentication protocols. Since machine learning methods are trained based on hidden concepts in biometric and physical layer data, these dynamic authentication models can be more reliable than traditional methods. The main advantage of these methods is that the behavioral traits of humans and devices are tough to counterfeit. Furthermore, machine learning facilitates continuous and context-aware authentication.

Advancements in digital communications that permit remote patient visits and condition monitoring can be attributed to a revolution in digital healthcare systems. Continuous authentication based on contextual information offers a number of advantages over traditional authentication, including the ability to estimate the likelihood that the users are who they claim to be on an ongoing basis over the course of an entire session, making it a much more effective security measure for proactively regulating authorized access to sensitive data. Current authentication models that rely on machine learning have their shortcomings, such as the difficulty in enrolling new users to the system or model training sensitivity to imbalanced datasets. To address these issues, we propose using ECG signals, which are easily accessible in digital healthcare systems, for authentication through an Ensemble Siamese Network (ESN) that can handle small changes in ECG signals. Adding preprocessing for feature extraction to this model can result in superior results. We trained this model on ECG-ID and PTB benchmark datasets, achieving 93.6% and 96.8% accuracy and 1.76% and 1.69% equal error rates, respectively. The combination of data availability, simplicity, and robustness makes it an ideal choice for smart healthcare and telehealth.

In federated learning, standard machine learning (ML) techniques are modified so they can be applied to data held by separate participants without the need for exchanging said data and while preserving privacy. Other data modelling techniques, such as singular value decomposition, have been similarly federated, enabling federated principal component analysis (PCA), which is a popular preprocessing step for ML tasks. Supervised PCA improves on standard PCA by using labeled data to retain more relevant information for supervised ML problems. However, a federated version of supervised PCA does not exist in the literature. In this paper, we propose a federated version of supervised PCA and its dual and kernel variations, called FeS-PCA, dual FeS-PCA, and FeSK-PCA, respectively. We used random orthogonal matrix masking to keep FeS-PCA and dual FeS-PCA private, while FeSK-PCA was kept private using an approximation of the standard approach. We tested our proposed approaches by recreating visualization, classification, and regression experiments from the original unfederated supervised PCA paper. We further added a real-world federated dataset to test the scalability and fidelity of our approach. Our analysis and results indicate that FeS-PCA and dual FeS-PCA are faithful, lossless, and private versions of their unfederated counterparts. Furthermore, despite being an approximation, FeSK-PCA achieves nearly identical performance to standard kernel SPCA in many cases. This is in addition to the added benefit of a reduced runtime and smaller memory footprint.

The action of spreading false information through fake news articles presents a significant danger to society because it has the ability to shape public opinion with inaccurate facts. This can lead to negative effects, such as reduced trust in institutions and the promotion of conflict, division, and even violence. In this article, a text augmentation technique is introduced as a means of generating new data from preexisting fake news datasets. This approach has the potential to enhance classifier performance by a range of 3%–11%. It can also be utilized to launch a successful attack on trained classifiers, with up to a 90% success rate. However, the success rate of these attacks decreased to less than 28% when the model was retrained with the generated adversarial examples. These results demonstrate the effectiveness of text augmentation as a viable method for detecting fake news and increasing classifier accuracy and performance, as well as its ability to be utilized to perform adversarial machine learning (ML) and improve the resilience of ML algorithms.