Updating A Heuristic And Optimizer Learning Model For Intrusion Detection In The Internet Of Things Environment

Abstract

Web technologies in the modern period have produced an enormous amount of data. The relationship between various gadgets and services has also been investigated to make effective and extensive use of modern technologies. Because of these restrictions, the probability of a security breach is significantly rising on devices with limited resources. Improved scalability and dependability of public services can be achieved by integrating an IoT backend with multi-cloud architecture. Managing user requests for IoT services may include several users accessing multi-cloud resources, posing a data security risk. Coming up with new functional components and security plans becomes more difficult. This study presents an advanced Intrusion Detection Model (IDM) designed to detect threats in both network and application contexts. The framework is organized into three essential phases: data preprocessing, feature selection, and classification. In the initial phase, the gathered datasets are subjected to a transformation process employing the Improved Rat Optimizer (IRO) method, which aims to achieve a balanced and standardized data representation. Following this, the feature selection process utilizes the Opposition Heuristic Learning (OHL), which mimics the adaptive behaviors exhibited by rats to systematically determine the most significant features for classification purposes. The stability of the selected features is ensured by evaluating their fittest value. In conclusion, the suggested model for binary classification is a Long Short Term Convolutional Network Model (LSTCNM), which utilizes a two-dimensional array structure to retain input features and handle intricate layer processing efficiently. The purpose of this model is to distinguish between typical and unusual network traffic. The suggested framework achieves  2.5% false positive rate, 97.24% detection rate and 95.20% accuracy when trained and assessed using Netflow-based datasets.