In today's industrial landscape, companies are increasingly using smart devices, sensors, and connected systems to make operations more efficient and productive. In this scenario, the risk of cyber threats becomes a major concern. Industrial Internet of Things (IIoT) systems play a crucial role in overseeing various aspects of industrial processes, spanning manufacturing, logistics, and energy management. Ensuring the security of these connected devices is essential to protect against potential cyberattacks that could disrupt operations, compromise sensitive data, or pose safety risks. If there is a breach in IIoT security, the consequences could be severe, leading to financial losses, harm to reputation, and, in specific industries, potential danger to human lives. The objectives of this study are the following: (i) designing advanced model for malicious access detection that takes advantage of multiple model’s prediction and (ii) optimizing its hyperparameters using metaheuristic optimization. In this work, a data-driven IIoT security framework has been designed by using AdaBoost-based model. Further, gravitational interaction optimization, a metaheuristic optimization, has been used to optimize the AdaBoost’s parameters, in order to achieve better malicious access detection rate. The proposed approach is trained and validated with IIoT network-simulated dataset and found to efficient.