Cyberattacks, particularly those that take place in real time, will be able to target an increasing number of networked systems as more and more items connect to the Internet of items. While the system is operational, it is susceptible to intrusions that might have catastrophic consequences, such as the theft of sensitive information, the violation of personal privacy, or perhaps physical injury or even death. These outcomes are all possible while the system is operational. A mixed-methods research approach was required in order to fulfill the requirements for understanding the nature and scope of real-time assaults on IoT-powered cybersecurity infrastructure. The quantitative data that was utilized in this research came from an online survey of IoT security professionals as well as an analysis of publicly available information on IoT security incidents. For the purpose of gathering qualitative data, in-depth interviews with industry experts and specialists in the area of Internet of Things security were conducted. The authors provide a novel method for identifying cybersecurity flaws and breaches in cyber-physical systems, one that makes use of deep learning in conjunction with blockchain technology. This method has the potential to be quite useful. Their proposed technique compares and evaluates unsupervised and deep learning-based discriminative methods, in addition to introducing a generative adversarial network, in order to determine whether cyber threats are present in IICs networks that are powered by IoT. The results indicate an improvement in performance in terms of accuracy, reliability, and efficiency in recognizing all types of attacks. The dropout value was found to be 0.2, and the epoch value was set at 25.