Artificial Intelligence for Cyber-Physical Systems Hardening

Preface
Contents
Introduction
1 Context and Definition
2 Characteristics and Design Goals
3 Security and Hardening
4 Intelligence
5 Summary
References
Machine Learning Construction: Implications to Cybersecurity
1 Introduction
1.1 Motivation
1.2 Notation
1.3 Roadmap
2 Statistical Decision Theory
2.1 Regression
2.2 Classification
2.3 Where Is Learning?
3 Parametric Regression and Classification
3.1 Linear Models (LM)
3.2 Generalized Linear Models (GLM)
3.3 Nonlinear Models
4 Nonparametric Regression and Classification
4.1 Smoothing Techniques
4.2 Additive Models (AM)
4.3 Generalized Additive Models (GAM)
4.4 Projection Pursuit Regression (PPR)
4.5 Neural Networks (NN)
5 Optimization
5.1 Introduction
5.2 Connection to Machine Learning
5.3 Types of MOP
6 Performance
6.1 Error Components
6.2 Receiver Operating Characteristic (ROC) Curve
6.3 The True Performance Is A Random Variable!
6.4 Bias-Variance Decomposition
6.5 Curse of Dimensionality
6.6 Performance of Unsupervised Learning
6.7 Classifier Calibration
7 Discussion and Conclusion
References
Machine Learning Assessment: Implications to Cybersecurity
1 Introduction
1.1 Motivation
1.2 Notation
1.3 Roadmap
2 Nonparametric Methods for Estimating the Bias and the Variance of a Statistic
2.1 Bootstrap Estimate
2.2 Jackknife Estimate
2.3 Bootstrap Versus Jackknife
2.4 Influence Function, Infinitesimal Jackknife, and Estimate of Variance
3 Nonparametric Methods for Estimating the Error Rate of a Classification Rule
3.1 Apparent Error
3.2 Cross Validation (CV)
3.3 Bootstrap Methods for Error Rate Estimation
3.4 Estimating the Standard Error of Error Rate Estimators
4 Nonparametric Methods for Estimating the AUC of a Classification Rule
4.1 Construction of Nonparametric Estimators for AUC
4.2 The Leave-Pair-Out Boostrap (LPOB) ModifyingAbove upper A upper U upper C With caret Super Subscript left parenthesis 1 comma 1 right parenthesisAUC"0362AUC( 1,1) , Its Smoothness and Variance Estimation
4.3 Estimating the Standard Error of AUC Estimators
5 Illustrative Numerical Examples
5.1 Error Rate Estimation
5.2 AUC Estimation
5.3 Components of Variance and Weak Correlation
5.4 Two Competing Classifiers
6 Discussion and Conclusion
7 Appendix
7.1 Proofs
7.2 More on Influence Function (IF)
7.3 ML in Other Fields
References
A Collection of Datasets for Intrusion Detection in MIL-STD-1553 Platforms
1 Introduction
2 Mil-STD-1553 Baseline
2.1 Major Components
2.2 Bus Communication
3 Mil-Std-1553 Attack Vectors
3.1 Assumptions and Attacker Position/foothold on 1553 Platform
3.2 Attack Vectors and Types
4 Simulation and IDS Dataset Generation
4.1 Simulation Setup
4.2 Baseline Scenarios and Datasets
4.3 Attack Scenarios and Datasets
5 Conclusion
References
Unsupervised Anomaly Detection for MIL-STD-1553 Avionic Platforms Using CUSUM
1 Introduction
2 Datasets
3 Features Model
4 Detection Model
4.1 Change Point Detection
4.2 Using CUSUM Algorithm
5 Empirical Evaluation
5.1 Performance Metrics
5.2 Evaluation Procedures
5.3 Evaluation Results
5.4 Results Discussion
6 Conclusion
References
Secure Design of Cyber-Physical Systems at the Radio Frequency Level: Machine and Deep Learning-Driven Approaches, Challenges and Opportunities
1 Introduction to Cyber-Physical Systems
1.1 Security and Application Areas of CPS
2 Critical Infrastructures
3 Fundamentals of Radio Frequency Fingerprinting
4 Application Areas of RF Security
4.1 Authentication
4.2 Geo-location and Tracking
4.3 Intrusion Detection
4.4 Interference Detection and Traditional Approaches
4.5 Anti-spoofing Solutions
5 Machine and Deep Learning-Based RF Security for Cyber-Physical Systems
5.1 Machine Learning Based Approaches for RF Fingerprinting
5.2 Proactive, Adaptive Machine Learning Based Approaches for RF Interference Detection
5.3 Machine Learning Based Anti-spoofing Approaches for RF Security
5.4 Machine Learning Based Antijamming Approaches for RF Security
6 Open Issues, Challenges and Opportunities in Secure Cyber-Physical Systems via RF Fingerprinting
6.1 Impact of Receiver Hardware
6.2 Robustness in Realistic Operation Environments
6.3 Simulation-Reality Gap
6.4 Finding a Realistic Dataset
6.5 Feature Selection
6.6 Other Open Issues
7 Summary
References
Attack Detection by Using Deep Learning for Cyber-Physical System
1 Introduction
2 CPSs and DLs
3 Different DL Models in CPSs
3.1 Convolutional Neural Networks (CNNs)
3.2 Auto Encoder (AE)
3.3 Deep Belief Network (DBN)
3.4 Recurrent Neural Network (RNN)
4 Leveraging DL to Detect Attacks in CPSs
4.1 Using Convolutional Neural Networks (CNNs)
4.2 Using Auto Encoder (AE)
4.3 Using Deep Belief Network (DBN)
4.4 Using Recurrent Neural Network (RNN)
5 Leveraging RL and DL in Detecting Cyberattacks in CPS
5.1 Using Reinforcement Learning (RL)
5.2 Deep Reinforcement Learning (DRL)
6 Data Acquisition in CPSs
7 Challenges to Attack Detection in CPSs
8 Robust Attacks Detection
9 Conclusion
References
Security and Privacy of IoT Devices for Aging in Place
1 Introduction
2 Review of IOT Devices for Aging in Place
2.1 Device Types
2.2 Use Cases
3 AIP Threats and Vulnerabilities
3.1 Threats Speficic to AgeTech Environment
3.2 Common Threats
3.3 Device Specific Threats
3.4 Mitigation Strategies
4 Using Machine Learning and AI Models
4.1 Available Datasets
4.2 Existing ML-Based Proposals
5 Conclusion
References
Detecting Malicious Attacks Using Principal Component Analysis in Medical Cyber-Physical Systems
1 Introduction
2 Concepts of Intrusion Detection Systems
3 PCA Based Anomaly Detection
4 Experimental Evaluations
5 Conclusions and Future Work
References
Activity and Event Network Graph and Application to Cyber-Physical Security
1 Introduction
2 AEN Graph Theoretic Model
3 AEN Data Sources
4 Graph Model Elements
4.1 AEN Nodes
4.2 AEN Edges
5 AEN Probability Model
5.1 Probability Model Definition
5.2 Probability Model Usage and Application
6 Graph Construction and Framework Implementation
6.1 Framework Architecture
6.2 Case Study Based on a Cyperphysical Security Dataset
7 Conclusion
References