Human action recognition in industrial manufacturing (HAR-IM) plays a crucial role in automating the identification and classification of worker activities. These tasks are inherently fine-grained, requiring the recognition of subtle movements and intricate tool interactions in complex and cluttered environments. While recent advancements in skeleton- and RGB-based recognition methods have demonstrated success in general scenarios, their performance often declines in industrial settings due to challenges such as visually similar actions and significant background noise, which obscure critical motion cues. To address these issues, we propose FusionARM, a multimodal framework that harnesses the complementary strengths of skeleton and RGB modalities. FusionARM leverages skeleton data to prioritize salient spatio-temporal regions within RGB frames, effectively filtering redundant background information and improving the localization of fine-grained activities. The framework introduces a Spatial-Temporal Relevance Selection (STR) mechanism, which aligns keyframes and joint positions to ensure effective fusion of skeleton dynamics with visual context. Additionally, a Model Fusion strategy adaptively balances the contributions of each modality, generating representations that are both discriminative and noise-resilient. Extensive experiments on HAR-IM benchmark datasets validate that FusionARM outperforms state-of-the-art methods, demonstrating its effectiveness in tackling the unique challenges posed by industrial environments.

Unknown industrial control protocols (ICPs) seriously hamper the device intercommunication and security analysis of the Industrial Internet of Things due to the absence of public specification information. Protocol reverse analysis has emerged as a promising technology to infer their specifications, where the primary step is to extract protocol fields by locating their boundaries in the network packet. Previous works leverage various algorithms, such as sequence alignment, keyword mining, and statistic analysis for field extraction. However, they have limitations in excavating the unique features of ICP fields, leading to inaccuracies in boundary localization. To address this problem, we propose an entropy-based field segmentation method for unknown ICPs. After stacking protocol packets vertically, we calculate the information entropy and information gain ratio of data values at each location in the packet. By analyzing the distribution variations of these entropy features in diverse ICP fields, we derive multiple packet segmentation rules to locate the field boundaries. Extensive comparative experiments demonstrate the superiority of our method for ICP field extraction.

Sharding is widely regarded as a highly promising solution to address the scalability limitations of blockchain. However, the scalability and throughput improved by using sharding are limited by the verification of cross-shard transactions. To reduce cross-shard transaction and improve the throughput of the blockchain, the existing sharding schemes are based on factors such as the edge-end structure of Industrial Internet of Things (IIoT) for sharding. But these schemes are centralized, leading to the problems of low sharding efficiency, poor scalability, and poor security. Moreover, these schemes adopt a nonoverlapping sharding architecture, so the verification cost of cross-shard transactions is significantly higher than that of intrashard transactions. In order to solve the above problems, this article proposes an overlapping self-organizing sharding scheme (deep reinforcement learning (DRL)-OSS) for large-scale IIoT blockchain. Based on local blockchain information, such as nodes’ information and transaction interaction frequency, DRL-OSS uses DRL to achieve self-organizing sharding with the aim to maximize the throughput and security of blockchain. In addition, based on the threat model, this article also designs a block complaint scheme (BCS) to further improve the security of the blockchain, thereby avoiding the reduction in resistance to 1% attack due to the poor anti-predictability of shards and the dilution of computing power. Through experimental verification and analysis, DRL-OSS improves the throughput by 50% when compared to state-of-the-art sharding schemes and has higher system security.