ERROR RATES, TIMESTAMP, AND FLOW DURATION ANALYSIS-BASED ATTACK PATTERN IDENTIFICATION FRAMEWORK USING EAE-DBSCAN
Amaresan Venkatesan · Journal of Artificial Intelligence Machine Learning and Data Science · 2023
For ensuring Network Security (NS) against malicious activities, Dedicated Link Aggregation (DLA) in Computer Network Traffic (CNT) optimizes data transmission with increased bandwidth and reliability.Nevertheless, the traditional works failed to identify the Attack Patterns (APs) centred on timestamps, Error Rates (ERs), and flow duration, thereby resulting in inefficiencies in threat detection in NS.Thus, this paper proposes Ensemble Adaptive Entropy Density-Based Spatial Clustering of Applications with Noise (EAE-DBSCAN) and MeDecay Heuristic-based Radial Basis Function Networks (MDH-RBFN) techniques to identify patterns and classify the normal and malicious traffic, respectively.Primarily, the data is pre-processed, followed by DLA utilizing EAE-DBSCAN and feature extraction.After that, by using EAE-DBSCAN, the patterns are identified from the extracted features for enhanced network performance.Subsequently, utilizing MDH-RBFN, the data is categorized as normal and malicious traffic with a Mean Absolute Error (MAE) of 0.0025.Here, the malicious traffic is blocked, whereas the non-attacked data is encrypted.Thereafter, the traffic level is predicted for non-attacked traffic data as low, medium, high, very high, and extreme.At last, the required loads are balanced for storing data in the cloud.