The interconnected and diverse nature of Digital Twin (DT)-based industrial Artificial Internet of Things (AIoT) systems exposes them to potential cyber threats and malicious activities. This paper introduces a novel framework called BLIND-Twin, which leverages blockchain, DT, and Large Language Model (LLM) technologies to address critical security and scalability challenges in industrial AIoT networks. By integrating DT technology, BLIND-Twin continuously mirrors physical environments, enabling real-time monitoring and synthetic data generation to simulate diverse threat scenarios. Data from the DT undergoes feature extraction via a Long Short-Term Memory (LSTM) Autoencoder (LSTM-AE) to extract essential temporal patterns, while the LLM enables adaptive, contextaware intrusion detection without retraining. A permissioned blockchain layer ensures data integrity, privacy, and secure logging through smart contracts, supporting automated threat response with verifiable audit trails. The framework's decentralized architecture mitigates Single Points of Failure (SPoF), addressing scalability and privacy concerns. Performance evaluations utilizing datasets like 5G-NIDD and CICIoT2023 demonstrate BLINDTwin's capability in accurately detecting various cyber threats by achieving 99.63 % accuracy with minimal latency, showcasing its effectiveness for complex industrial AIoT environments.

The advancement of intelligent transport systems and the rise of autonomous vehicles offer significant potential for reducing road accidents. However, mountainous regions, such as South Korea, are particularly susceptible to the formation of black ice, which poses a serious risk to both human and autonomous drivers due to its near-invisibility and sudden formation. Traditional road condition monitoring methods often fail to promptly detect black ice, underscoring the need for more advanced sensing systems. This work presents a multimodal system called BlackIceNet, integrating visual, acoustic, and sensor data, including surface and ambient temperatures, to detect black ice. The system utilizes a convolutional neural network (CNN)-based framework for image and audio analysis, followed by data fusion techniques to combine the insights from each modality. The proposed algorithm includes the following steps: preprocessing and normalizing data, feature extraction from visual and acoustic data, multimodal fusion to combine vision, audio, and sensor data, and classification of road surface conditions using BlackIceNet. The dataset has been gathered over the years from various testbeds established across three distinct areas in South Korea, contributing to a comprehensive understanding of the area’s diverse conditions. The evaluation results demonstrate the efficacy of the proposed method, achieving a 97.54% accuracy rate in detecting black ice with a model size of 233.47 MB and a training time of 2293.92 s, offering a more compact and computationally efficient solution. This fusion-based approach overcomes the limitations of individual modalities, providing reliable and early warnings, thereby enhancing road safety in hazardous conditions.

The immersive metaverse environment offers distinct social interactions and opportunities, yet it also presents significant challenges in securely managing misbehavior, including hate speech, bullying, and harassment. Existing solutions primarily focus on detecting such behavior through artificial intelligence but lack robust mechanisms for management and governance. This gap is critical as the metaverse continues to mirror complex real-world interactions and centralized authority systems prove vulnerable to compromise. Our research introduces a novel framework, Meta-Governance, which not only detects but also effectively manages and governs user behavior through smart contracts, ensuring a secure, fair, and transparent metaverse environment. The system incorporates behavior monitoring to identify and condemn inappropriate behavior, specifically targeting problems such as hate speech and cyberbullying. Occurrences of misbehavior are permanently preserved on the blockchain to ensure the capacity to trace and bear accountability. In this article, we deploy a Natural Language Processing (NLP) model and a smart contract-based framework to address unusual behavior monitoring, access control, and credit scoring. Deep learning models are used to identify and classify linguistic patterns that may be considered hazardous. Blockchain technology addresses virtual misconduct using smart contracts, while a distinctive credit scoring mechanism ensures that users are held responsible for making disrespectful statements. The efficacy of the proposed smart contract is comprehensively evaluated within the context of a private Hyperledger Besu system. The integration of AI and blockchain may greatly improve the security and inclusiveness of the metaverse, highlighting the crucial role of these technologies in combating hate speech and enhancing user engagement.

Unmanned aerial vehicle (UAV) contributes substantial strategic benefits on the Internet of Military Things (IoMT). However, the untrusted party’s misuse of the UAV may violate the security and even demolish the critical operation in the IoMT system. In addition, data manipulation and falsification using unauthorized access are the significant challenges of the IoMT system. In response to this problem, this study proposes a blockchain-integrated convolution neural network (CNN)-based intelligent framework named IoMT-Net for identification and tracking illegal UAV in the IoMT system. Blockchain technology prevents illicit access, data manipulation, and illegal intrusions, as well as stored data on the central control server (CCS). Concurrently, the proposed CNN analyzed the radio-frequency (RF) signal sent by the antenna array element to determine the Direction of Arrival (DoA) for the localization of the illegal UAV. Therefore, a signal model is designed to process the received signal array through IoMT-Net. Moreover, the proposed CNN model is designed with two different functional modules, such as the resource accuracy tradeoff (RAT) module and the unique feature extraction and accuracy boosting (UAB) module, by adopting depthwise and grouped convolution layers. These sparsely connected convolution layers offer high DoA estimation accuracy while maintaining the network more lightweight. In addition, the skip connection is also leveraged into the subunits of RAT and UAB modules for sharing features and handling the vanishing gradients problem. Based on the simulation results, the proposed network achieves superior DoA estimation accuracy (approximately 97.63% accuracy at 10-dB SNR) and outperforms other state-of-the-art models.

Solar irradiance prediction is an indispensable area of the photovoltaic (PV) power management system. However, PV management may be subject to severe penalties due to the unsteadiness pattern of PV output power that depends on solar radiation. A high-precision long short-term memory (LSTM)-based neural network model named SIPNet to predict solar irradiance in a short time interval is proposed to overcome this problem. Solar radiation depends on the environmental sensing of meteorological information such as temperature, pressure, humidity, wind speed, and direction, which are different dimensions in measurement. LSTM neural network can concurrently learn the spatiotemporal of multivariate input features via various logistic gates. Moreover, SIPNet can estimate the future solar irradiance given the historical observation of the meteorological information and the radiation data. The SIPNet model is simulated and compared with the actual and predicted data series and evaluated by the mean absolute error (MAE), mean square error (MSE), and root MSE. The empirical results show that the value of MAE, MSE, and root mean square error of SIPNet is 0.0413, 0.0033, and 0.057, respectively, which demonstrate the effectiveness of SIPNet and outperforms other existing models.