Contents

• Dedication v
• Acknowledgements vii
• Preface xv
• 1 Introduction 1
  • 1.1 Setting up your Environment 2
  • 1.1.1 Anaconda 2
  • 1.1.2 Setting up your Environment with a Physical Box 2
  • 1.2 Background 3
  • 1.3 Companion GitHub Code and YouTube videos 3
  • 1.4 Numpy Arrays, Tensors, and Linear Algebra 4
  • 1.4.1 Numpy Arrays 5
  • 1.4.2 Tensor Operations with PyTorch 16
  • 1.5 Conclusion 20
• 2 Introduction to Cyber Security and Machine Learning 23
  • 2.1 Risk 23
  • 2.1.1 Expected Risk given Threat and Impact Probabilities 23
  • 2.1.2 Entropy 25
  • 2.2 Computer Hacking 26
  • 2.3 Network Security 27
  • 2.4 Software Security and Malware 29
  • 2.5 Social Engineering 34
  • 2.6 Cryptography 35
  • 2.7 AI Assurance 36
  • 2.7.1 AI Auditing and Explainability 36
  • 2.7.2 Bias Testing 37
  • 2.7.3 Adversarial Attacks 37
  • 2.8 Introduction to Cyber Security and Machine Learning Literature 37
  • 2.9 Summary 38
• 3 Traditional Machine Learning 39
  • 3.1 Code Issues 40
  • 3.2 Object Oriented Programming 43
  • 3.3 Performance Evaluation 44
  • 3.3.1 Regression Performance Evaluation 44
  • 3.3.2 Classification Performance Evaluation 45
  • 3.3.3 Plotting Performance 47
  • 3.4 Information Theory 49
  • 3.4.1 Entropy 49
  • 3.4.2 Kullback-Leibler (KL) divergence 52
  • 3.4.3 Mutual Information 53
  • 3.4.4 Information Gain 55
  • 3.4.5 Conditional Entropy 56
  • 3.5 Optimization 60
  • 3.6 Anomaly Detection Algorithms 63
  • 3.6.1 Hopfield Networks 63
  • 3.6.2 Boltzman Machines 67
  • 3.7 Popular Supervised Learning Algorithms 67
  • 3.7.1 KNN 67
  • 3.7.2 Logistic Regression 73
  • 3.7.3 Neural Networks 74
  • 3.7.4 Regression, Trees, and XGBoost 75
  • 3.8 Summary 76
• 4 Data Loading and Pre-processing 77
  • 4.1 Loading the Data 77
  • 4.2 Data and Feature Pre-Processing 80
  • 4.2.1 CSV Files 80
  • 4.3 One Hot Encoding 81
  • 4.4 Features 83
  • 4.4.1 Sniffing and Spoofing 85
  • 4.4.2 Features from Network Data 88
  • 4.4.3 Features from Malware 96
  • 4.4.4 Features from Forensics 102
  • 4.4.5 Features from Text 103
  • 4.4.6 Features from Websites for Web Security 105
  • 4.4.7 Features from Images 106
  • 4.5 Corpora (i.e. Annotated Datasets) 112
  • 4.6 Summary 114
• 5 Deep Learning: Starting at the beginning 115
  • 5.1 Things to know about the code 115
  • 5.2 Getting Started with Deep Leaning 118
  • 5.3 Deep Leaning Definition 121
  • 5.4 PyTorch Basics 121
  • 5.5 Loading your Data into your PyTorch code 123
  • 5.6 Linear Regression 127
  • 5.6.1 Theory and Intuition of Linear Regression 127
  • 5.6.2 Linear Regression model for Wine Quality data 131
  • 5.7 Linear Regression to Logistic Regression and Neural Networks 147
  • 5.7.1 Regression vs Classification 147
  • 5.7.2 Instantiating the different models 148
  • 5.7.3 Reading Data in batches 149
  • 5.7.4 What is Number of Epochs? 150
  • 5.8 Logistic Regression 151
  • 5.8.1 Theory and Intuition of Logistic Regression 151
  • 5.8.2 Entropy and Cross Entropy 158
  • 5.8.3 Logistic Regression NN Architecture 171
  • 5.9 Layers of the Neural Network in PyTorch 173
  • 5.9.1 Intuition of adding more layers to a Neural Network 175
  • 5.10 Going Deep: An N layer Neural Network in PyTorch 175
  • 5.11 Easy Deep Learning with the Iris dataset 178
  • 5.12 More Challenging Deep Learning with the Wine Quality dataset 192
  • 5.13 Transfer Learning 204
  • 5.13.1 Important Ideas in Transfer Learning 204
  • 5.13.2 Pre-training 204
  • 5.13.3 Fine-tuning 205
  • 5.13.4 Freezing some weights of the neural network 205
  • 5.13.5 Sharing the weights from one model to another 207
  • 5.13.6 Reinforcement Learning Through Human Feedbacks (RLHF) 207
  • 5.13.7 Transfer Learning Performance Metrics 207
  • 5.14 Summary 208
• 6 Network Security 209
  • 6.1 Intrusion Detection Systems (IDS) 209
  • 6.2 Anomaly Detection with Packet Payloads 216
  • 6.3 Restricted Boltzman Machines 223
  • 6.4 Privacy Preserving Auto-Encoder 226
  • 6.5 Summary 230
• 7 Software Security and Malware 231
  • 7.1 Malware Detection 231
  • 7.2 Transfer Learning for Malware Detection 237
  • 7.3 Summary 249
• 8 Social Engineering 251
  • 8.1 Phishing Classifier Using Meta-Data 251
  • 8.2 Phishing Detection using Transfer Learning and Text Features 259
  • 8.3 Phishing Detection specific Tools such as NVIDIA’s Morpheus 261
  • 8.4 Summary 261
• 9 Cryptography 263
  • 9.1 Cracking Hashes 263
  • 9.2 Homomorphic Cryptography and Torch 269
  • 9.3 Summary 272
• 10 AI Assurance 273
  • 10.1 AI Auditing and Explainability 273
  • 10.1.1 Shapley 274
  • 10.2 Bias Testing 279
  • 10.2.1 The 4/5s metric 279
  • 10.2.2 Differential Validity Tests 280
  • 10.3 Adversarial Machine Learning 280
  • 10.3.1 A Concrete Adversarial ML Example with Images 284
  • 10.3.2 Attacks on LLMs with Pre-existing Tokens 302
  • 10.3.3 Adversarial Perturbation of Text Embeddings 303
  • 10.4 Software Testing 306
  • 10.5 Data or Model Drift 307
  • 10.6 AI Toxicity 309
  • 10.7 Federated Learning and Privacy 311
  • 10.8 Summary 316
• 11 Conclusions and Final Thoughts 317
  • 11.1 The Cyber Security and AI horizon 317
  • 11.2 Summary 318
• Other Books by Ricardo A. Calix 319
• The Author: Ricardo A. Calix 321
• Glossary 323
• Bibliography 325
• An Important Final Note 327