TOWARDS ROBUST CYBER DEFENSE: ENSEMBLE MACHINE LEARNING MODELS FOR DARKNET TRAFFIC CLASSIFICATION

Cyber intelligence groups occasionally bring up the "darknet," a section of the internet that most users do not think is suitable for M2M communication. Evaluating the possible risks to the network must precede any initiatives to enhance its defenses. This study introduces a unique method for machine learning classification called stacking ensemble learning, aimed at assessing and categorizing darknet data. This study attained a 98% accuracy rate in differentiating between Darknet and benign traffic by using the newly available CIC-Darknet2020 dataset and ensembles of machine learning algorithms. It had a 97% success rate in accurately identifying the specific program type accountable for the Darknet traffic. We employed a game-theory-based method to evaluate the results of the machine learning models and demonstrate the influence of the features in order to deepen our comprehension of Darknet traffic patterns. To the best of everyone's knowledge, the dataset's creators have attested to the unprecedented nature of our analysis.