Internet of Things (IoT) devices are leading to advancements in innovation, efficiency, and sustainability across various industries. However, as the number of connected IoT devices increases, the risk of intrusion becomes a major concern in IoT security. To prevent intrusions, it is crucial to implement intrusion detection systems (IDSs) that can detect and prevent such attacks. IDSs are a critical component of cybersecurity infrastructure. They are designed to detect and respond to malicious activities within a network or system. Traditional IDS methods rely on predefined signatures or rules to identify known threats, but these techniques may struggle to detect novel or sophisticated attacks. The implementation of IDSs with machine learning (ML) and deep learning (DL) techniques has been proposed to improve IDSs’ ability to detect attacks. This will enhance overall cybersecurity posture and resilience. However, ML and DL techniques face several issues that may impact the models’ performance and effectiveness, such as overfitting and the effects of unimportant features on finding meaningful patterns. To ensure better performance and reliability of machine learning models in IDSs when dealing with new and unseen threats, the models need to be optimized. This can be done by addressing overfitting and implementing feature selection. In this paper, we propose a scheme to optimize IoT intrusion detection by using class balancing and feature selection for preprocessing. We evaluated the experiment on the UNSW-NB15 dataset and the NSL-KD dataset by implementing two different ensemble models: one using a support vector machine (SVM) with bagging and another using long short-term memory (LSTM) with stacking. The results of the performance and the confusion matrix show that the LSTM stacking with analysis of variance (ANOVA) feature selection model is a superior model for classifying network attacks. It has remarkable accuracies of $96.92\%$ and $99.77\%$ and overfitting values of $0.33\%$ and $0.04\%$ on the two datasets, respectively. The model’s ROC is also shaped with a sharp bend, with AUC values of $0.9665$ and $0.9971$ for the UNSW-NB15 dataset and the NSL-KD dataset, respectively.

This paper proposes a highly secure image encryption technique based on chaotic synchronization. Firstly, through the design of a synchronization controller, we ensure that the master–slave chaotic extremum-coded random number generators (ECRNGs) embedded in separated transmitters and receivers are fully synchronized to provide synchronized dynamic random sequences for image encryption applications. Next, combining these synchronized chaotic sequences with the AES encryption algorithm, we propose an image segmentation and multi-encryption method to enhance the security of encrypted images and realize a secure image transmission system. Notably, in the design of the synchronization controller, the transient time before complete synchronization between the master and slave ECRNGs is effectively controlled by specifying the eigenvalues of the matrix in the synchronization error dynamics. Research results in this paper also show that complete synchronization of ECRNGs can be achieved within a single sampling time, which significantly contributes to the time efficiency of the image transmission system. As for the image encryption technique, we propose the method of image segmentation and use the synchronized dynamic random sequences generated by the ECRNGs to produce the keys and initialization vectors (IVs) required for AES-CBC image encryption, greatly enhancing the security of the encrypted images. To highlight the contribution of the proposed segmented image encryption, statistical analyses are conducted on the encrypted images, including histogram analysis (HA), information entropy (IE), correlation coefficient analysis (CCA), number of pixels change rate (NPCR), and unified average changing intensity (UACI), and compared with existing literature. The comparative results fully demonstrate that the proposed encryption method significantly enhances image encryption performance. Finally, under the network transmission control protocol (TCP), the synchronization of ECRNGs, dynamic keys, and IVs is implemented as well as segmented image encryption and transmission, and a highly secure image transmission system is realized to validate the practicality and feasibility of our design.

In this paper, we propose an extracting method of pseudorandom bits from sequences generated by Shirase's Me operation on elliptic curves over finite fields, whose security is based on the hardness of the Elliptic Curve Discrete Logarithm Problem (ECDLP). In Shirase's original pseudorandom number generator, only the least significant bit (LSB) of the $x$ coordinate of each point generated by the Me operation was extracted. However, since this approach is time-consuming and impractical for real-world use, we conclude a new method that extends the number of extracted bits. We evaluate the statistical properties of the generated sequences from multiple perspectives: 0/1 occurrence rate at each bit position, randomness with respect to initial values, calculation cost, and randomness of the $y$-coordinates. Furthermore, we compared the ECDLP based pseudorandom number generator with the DLP based method on which it is derived, in order to demonstrate the superiority of the ECDLP based approach. The comparison was conducted from three perspectives: 0/1 occurrence rate at each bit position, randomness with respect to initial values, generation speed and memory usage. Experimental results confirmed that the ECDLP based method achieved higher randomness and stability across all metrics, demonstrating its superiority over the DLP based method. These findings suggest that the generator has strong potential for practical deployment in cryptographic applications requiring secure and efficient pseudorandom number generation.

IoT SecurityIn the Research Article (DOI: 10.1002/aisy.202501500), Hirofumi Tanaka and co-workers demonstrate a novel in-materio encryption method for IoT edge systems using Ag/Ag2S nanoparticle devi...

In Internet of Things (IoT) edge environments, securing sensor data with minimal computational overhead is increasingly important, as conventional software‐based cryptographic methods require significant computational resources and remain vulnerable to cyber‐physical attacks such as side‐channel attacks. In this study, we demonstrate an in‐materio lightweight cryptographic scheme that leverages the nonlinear response and short‐term memory of Ag/Ag ₂ S nanoparticle devices to directly encrypt analog signals without relying on any external computational resources. By sequentially processing electrocardiogram signals through a series of nanoparticle devices, the encrypted outputs exhibit enhanced complexity and statistical randomness and pass the NIST SP 800‐22 randomness tests. Decryption is performed using step‐by‐step inverse conversion models constructed with echo‐state networks, enabling the reconstruction of the original signal with 91% accuracy while preserving the intrinsic device dynamics. These results establish a computation resource‐free physical‐layer encryption approach based on nanomaterial behavior and highlight its potential as a practical and computationally efficient security mechanism for next‐generation IoT edge systems.