Anomaly detection (AD) in the industrial internet of things (IIoT) platform is said to be the major module of security the consistency, safety, and efficacy of industrial procedures. In IIoT, several sensors and devices always gather massive data counts from tools, infrastructure, and equipment. Ensuring the safety of IIoT AD systems is vital for protecting against cyber-attacks. AD and Intrusion detection are required to work in tandem to maintain industrial procedures. Machine learning (ML) and deep learning (DL) approaches generally employed for AD in IIoT. These methods allow automated AD and pattern detection. With this motivation, this study develops an enhanced pelican optimization model with an ensemble voting-based anomaly detection (EPOA-EVAD) approach. Primarily, the SMOTE approach is used for class imbalance data handling. The presented EPOA-EVAD technique integrates the EPOA-based feature selection and ensemble voting classifier to enhance the accuracy and efficiency of AD in IIoT environments. In addition, the group teaching optimization algorithm with SMOTE approach is used for the purpose of class imbalance data handling. EPOA optimally chooses the related features from the complex and high-dimensional IIoT data. For AD, the EPOA-EVAD technique involves an ensemble learning process comprising gated recurrent unit (GRU), bi-directional long short-term memory (BiLSTM), and stacked autoencoder (SAE). Lastly, the seagull optimization (SGO) algorithm fine-tunes the parameters compared with the DL approaches. The simulation values exhibited that the proposed system outperforms standard approaches, offering robust and adaptive AD capabilities in dynamic IIoT settings. The proposed EPOA-EVAD technique contributes to the advancement of AD techniques crucial for maintaining the integrity and efficiency of industrial processes in the era of IIoT.

工业物联网 (IIoT) 平台中的异常检测 (AD) 据说是确保工业流程的一致性、安全性和有效性的主要模块。在工业物联网中，多个传感器和设备总是从工具、基础设施和设备中收集大量数据。确保 IIoT AD 系统的安全对于防范网络攻击至关重要。AD 和入侵检测需要协同工作以维护工业流程。机器学习 (ML) 和深度学习 (DL) 方法通常用于 IIoT 中的 AD。这些方法允许自动 AD 和模式检测。出于这个动机，本研究开发了一种增强的鹈鹕优化模型，采用基于集成投票的异常检测（EPOA-EVAD）方法。SMOTE 方法主要用于类不平衡数据处理。所提出的 EPOA-EVAD 技术集成了基于 EPOA 的特征选择和集成投票分类器，以提高 IIoT 环境中 AD 的准确性和效率。此外，采用SMOTE方法的小组教学优化算法来处理班级不平衡数据。EPOA从复杂且高维的IIoT数据中最优地选择相关特征。对于 AD，EPOA-EVAD 技术涉及集成学习过程，包括门控循环单元 (GRU)、双向长短期记忆 (BiLSTM) 和堆叠自动编码器 (SAE)。最后，与深度学习方法相比，海鸥优化（SGO）算法对参数进行了微调。仿真值表明，所提出的系统优于标准方法，在动态 IIoT 设置中提供强大的自适应 AD 功能。所提出的 EPOA-EVAD 技术有助于 AD 技术的进步，这对于维持 IIoT 时代工业流程的完整性和效率至关重要。