The rapid development of the Internet and smart devices increases network traffic, which in turn makes the infrastructure increasingly complex and heterogeneous. Distributed networks that generate massive volumes of data daily include mobile phones, wearable technology, and autonomous vehicles. Intrusion detection systems considerably improve the security and privacy of such devices. Intrusion detection for such paradigms is a non-trivial challenge that has gained extra significance because of the rapid expansion in the number and variety of security threats for such systems. However, due to the unique characteristics of such systems, such as battery power, bandwidth and CPU overheads, and network dynamics, intrusion detection for the Industrial Internet of Things (IIoT) is a challenge that necessitates taking into account the trade-off between detection accuracy and performance overheads. The decentralized learning technique known as federated learning (FL), which trains models locally and communicates the parameters to a centralized server, is a suitable illustration. The current study aims to provide a comprehensive and in-depth investigation of the use of FL-based intrusion detection systems for IIoT.