Anomaly detection in network traffic is one of the key techniques to ensure security in future networks. Today, the importance of this topic is even higher, since the network traffic is growing and there is a need to have smart algorithms, which can automatically adapt to new network conditions, detect threats and recognize the type of the possible network attack. Nowadays, there are a lot of different approaches, some of them have reached relatively sufficient accuracy. However, the majority of works are being tested on old datasets, which do not reflect current network conditions and it leads to overfitted results. This is caused by high redundancy of the data and because they fail to reflect the performance of the latest methods in the real-world anomaly detection applications. In this work, we applied a couple of new methods based on convolutional neural networks: U-Net based and Temporal convolutional network based for network attack classification. We trained and evaluated methods on the old dataset KDD99 and the modern large-scale one CSE-CIC-IDS2018. According to results, Temporal convolutional network with LSTM has achieved accuracy 92% and 97% on the KDD99 and the CSE-CIC-IDS2018 respectively, the U-Net model has accuracy 93% and 94% on the KDD99 and the CSE-CIC-IDS2018 respectively. Additionally, we utilized the focal loss function in the Temporal convolutional network with Long Short-Term Memory model, which has positive effect on class imbalance in time-series data. We showed, that the Temporal convolutional network in combination with Long Short-Term Memory network and U-Net model can give higher accuracy compared to other network architectures for network traffic classification. In this work we also proved, that methods trained on the old dataset can easily overfit during training and achieve relatively good results on the testing set, but at the same time, these methods are not so successful on more complex and actual data.

Remote continuous patient monitoring is an essential feature of eHealth systems, offering opportunities for personalized care. Among its emerging applications, emotion and stress recognition hold significant promise, but face major challenges due to the subjective nature of emotions and the complexity of collecting and interpreting related data. This paper presents a review of open access multimodal datasets used in emotion and stress detection. It focuses on dataset characteristics, acquisition methods, and classification challenges, with attention to physiological signals captured by wearable devices, as well as advanced processing methods of these data. The findings show notable advances in data collection and algorithm development, but limitations remain, e.g., variability in real-world conditions, individual differences in emotional responses, and difficulties in objectively validating emotional states. The inclusion of self-reported and contextual data can enhance model performance, yet lacks consistency and reliability. Further barriers include privacy concerns, annotation of long-term data, and ensuring robustness in uncontrolled environments. By analyzing the current landscape and highlighting key gaps, this study contributes a foundation for future work in emotion recognition. Progress in the field will require privacy-preserving data strategies and interdisciplinary collaboration to develop reliable, scalable systems. These advances can enable broader adoption of emotion-aware technologies in eHealth and beyond.

The Internet of Medical Things (IoMT) transforms healthcare by allowing real-time monitoring, diagnosis, and treatment using interconnected medical devices and sensors. However, the rapid growth of IoMT brings significant security and privacy challenges due to its critical vulnerability to cyber-attacks. This paper introduces a novel deep learning approach designed to analyze IoMT traffic and identify malicious activities. By leveraging the CIC IoMT dataset, we improved an existing neural network model to improve prediction accuracy. Our approach combines UNet++ and Long Short-Term Memory (LSTM) models to extract network traffic features effectively. Experimental results show that the proposed model outperforms traditional algorithms, achieving an accuracy of 99.92% in anomaly detection and 87.96% in attack categorization. Finally, we highlight the main limitations as well as possibilities for real-world implementation of the approach.

The intersection of technological advancements and cultural heritage studies has intensified the exploration of historical treasures, captivating historians and enthusiasts alike. Artificial intelligence now plays a key role in forensic art investigations by uncovering hidden patterns to detect forgeries. This study focuses on anomaly detection in X-ray images of paintings using the Ghent Altarpiece for training and testing purposes. We propose a novel model combining a Discriminatively Trained Reconstruction Anomaly Embedding Model (DRAEM), a Nested U-Net, and a new dataset derived from the Altarpiece. The proposed architecture was benchmarked against several state-of-the-art deep learning techniques in anomaly detection. Our model achieved an accuracy of 0.8399 and an F1 score of 0.7869, outperforming other methods in both accuracy and computational efficiency. Results, validated by a domain expert, show strong precision and computational efficiency through semi-supervised learning.