Cybersecurity is a significant obstacle in the modern society. As a result of the rapid advancement of technology, the world is becoming increasingly reliant on new technologies than ever before. The advancement is not only improving the living standards and allowing civilization to flourish, but also attracting a group of people to invent more and more sophisticated attacking techniques. As a consequence, it may be rather difficult to identify, prevent, or mitigate security risks and to manage them swiftly and efficiently. The need for regular updates to threat detection techniques arises from the increased frequency and emergence of new attacks. Conventional signature-based methods are unable to identify an unknown (new) attack, while they do their job quite efficiently against common ones. Because of this, machine learning (ML)-based intrusion detection systems (IDS) are becoming a more popular option. However, when a large number of features are present, it takes huge time to identify an attack. This paper presents a unique approach to reduce dimensionality of such data set using interval type-2 based fuzzy-rough sets. The technique is centered on removing redundancy from the reduced feature set by maximizing each feature’s relevance and significance. Modern routers are constructed with more configuration options, thus it is possible to add computer code to create a custom firmware that can be flashed into the router’s ROM so that the attack can be detected automatically. Due to lower complexity of the proposed algorithm, the integration with the router allows it to be effective in reducing system overhead and expediting the detection process. Exhaustive results are presented utilizing the commonly used classifiers to illustrate the efficacy of the suggested strategy on a real-world data set, along with a comparison with the other detection techniques.