Implementation of Machine Learning in Network Security

Introduction

ML is the combination of mathematics and statistics. ML algorithms identify patterns in the data from real time. By viewing this it is possible to identify correlations and also to detect anomalies. This chapter provides intuition to the reader about ML and its significant role in the network security. The reader could have basics of ML and design framework for security issues and have the practical skill to implement the ML algorithms. ML is for learning by experiences. It brings intelligence and adaptable to dynamic data through past data. Being get features of available data and relations among them, it could be possible to predict future outcomes. This could use these tasks: Information mining, and brings the inference and insight from data. Whenever we want to classify the elements, we first extract features from the elements and significant towards the label, which could differentiate between classes. We applied ML algorithms over the labeled data and model parameters are identified. These learned parameters are used to differentiate the unseen data, in the validation phase.

ML seems to be the best solution to cyber security in the large amount of digital era. This chapter focuses on little basics of ML and security. It picturizes the concept of ML in security and it is explained through use cases. The example is going to help the readers to get deep dive into the hands-on part of ML algorithms. It also gives the over view of ML and security state of the art. Key word in this technology world are Artificial intelligence (AI), ML and Deep Learning (DL) given in Figure 1.

Figure 1.

AI is the roof; Making things smart by machines and doing human like activities with intelligence. ML is the subset of AI and it learns from experience. ML is having capacity to recognize pattern without program. ML makes decisions based on data not from algorithm.

DL is the techniques for doing ML logics. DL recognizes patterns of pattern.

• •

  Learning [supervised, unsupervised, semi-supervised]

• •

  Supervised (Task driven approach): Classifier, Regression(prediction)

• •

  Unsupervised (Data Driven approach): Clustering

• •

  Semi supervised: combined both supervised and unsupervised

• •

  Ensemble Learning: Extension of Supervised learning. Combining simple models to get best output.

• •

  Reinforcement learning (Environment driven approach): Behavior react on changing environment based on rewards and punishments.

• •

  Active learning: Subset of Reinforcement learning. With teacher correction is possible other than environment changes

Ml And Security- State Of The Art

This section elaborates with current scenario of ML in security. Current trend in research area focused with three major areas such as security issues in network, security issues in IoT and security issues in medicine. These issues are resolved by researchers using several techniques. This chapter concentrates on ML techniques to resolve security issues in these areas. It also discusses three use cases and suggests tool for handling security challenges.