Toward Building an Intelligent and Secure Network: An Internet Traffic Forecasting Perspective
Abstract
Internet traffic forecast is a crucial component for the proactive management of self-organizing networks (SON) to ensure better Quality of Service (QoS) and Quality of Experience (QoE). Given the volatile and random nature of traffic data, this forecasting influences strategic development and investment decisions in the Internet Service Provider (ISP) industry. Modern machine learning algorithms have shown potential in dealing with complex Internet traffic prediction tasks, yet challenges persist. This thesis systematically explores these issues over five empirical studies conducted in the past three years, focusing on four key research questions: How do outlier data samples impact prediction accuracy for both short-term and long-term forecasting? How can a denoising mechanism enhance prediction accuracy? How can robust machine learning models be built with limited data? How can out-of-distribution traffic data be used to improve the generalizability of prediction models? Based on extensive experiments, we propose a novel traffic forecast/prediction framework and associated models that integrate outlier management and noise reduction strategies, outperforming traditional machine learning models. Additionally, we suggest a transfer learning-based framework combined with a data augmentation technique to provide robust solutions with smaller datasets. Lastly, we propose a hybrid model with signal decomposition techniques to enhance model generalization for out-of-distribution data samples. We also brought the issue of cyber threats as part of our forecast research, acknowledging their substantial influence on traffic unpredictability and forecasting challenges. Our thesis presents a detailed exploration of cyber-attack detection, employing methods that have been validated using multiple benchmark datasets. Initially, we incorporated ensemble feature selection with ensemble classification to improve DDoS (Distributed Denial-of-Service) attack detection accuracy with minimal false alarms. Our research further introduces a stacking ensemble framework for classifying diverse forms of cyber-attacks. Proceeding further, we proposed a weighted voting mechanism for Android malware detection to secure Mobile Cyber-Physical Systems, which integrates the mobility of various smart devices to exchange information between physical and cyber systems. Lastly, we employed Generative Adversarial Networks for generating flow-based DDoS attacks in Internet of Things environments. By considering the impact of cyber-attacks on traffic volume and their challenges to traffic prediction, our research attempts to bridge the gap between traffic forecasting and cyber security, enhancing proactive management of networks and contributing to resilient and secure internet infrastructure.