Network Traffic Anomaly Detection using ML

Abstract

Networked computer systems are deeply integrated in every aspect of our information-overloaded modern society.The mechanisms that keep our modern society flowing smoothly, with activities such as efficient execution of government and commercial transactions and services, or consistent facilitation of social transactions among billions of users, are all dependent on large networked computer systems.Today, every aspect of our lives is influenced by networked computer systems.The Internet, which provides transportation to all types of information in-including complex realtime multi-media data, is the universal network of millions of interconnected computer systems, organized as a network of thousands of distinct smaller networks.The recent growth of the Internet has been phenomenal and consequently, the computers and the networks that make the Internet hum have become the targets of enemies and criminals.Intrusions into a computer or network system are activities that destabilize them by compromising security in terms of confidentiality, availability or integrity, the three main characteristics of a secure and stable system.Machine learning is used to extract valid, novel, potentially useful and meaningful patterns from a dataset, usually large, in a domain of interest by using non-trivial mechanisms.A machine learning algorithm attempts to recognize complex patterns in datasets to help make intelligent decisions or predictions when it encounters new or previously unseen data instances.To deal with unseen examples, a machine learning algorithm must be cognizant of this necessity and thus, when it learns it must make conscious and diligent efforts to generalize from examples it has seen.Good generalization from data is a prime activity a learner program must perform.