دانلود کتاب Data-Driven Modeling of Cyber-Physical Systems using Side-Channel Analysis

فهرست مطالب :

Preface
Acknowledgments
Contents
1 Introduction
1.1 Cyber-Physical System
1.2 Data-Driven Modeling
1.3 Side-Channel Analysis
1.4 Book Sections
1.4.1 Part I: Data-Driven Attack Modeling
1.4.2 Part II: Data-Driven Defense of Cyber-Physical Systems
1.4.3 Part III: Data-Driven Digital Twin Modeling
1.4.4 Part IV: Non-Euclidean Data-Driven Modeling of Cyber-Physical Systems
1.5 Summary
References
Part I Data-Driven Attack Modeling
2 Data-Driven Attack Modeling Using Acoustic Side-Channel
2.1 Introduction
2.1.1 Research Challenges and Contributions
2.2 Background and Related Work
2.3 Sources of Acoustic Emission
2.3.1 System Description
2.3.2 Equation of Motion
2.3.3 Natural Rotor Oscillation Frequency
2.3.4 Stator Natural Frequency
2.3.5 Source of Vibration
2.3.5.1 Electromagnetic Source
2.3.5.2 Mechanical Source
2.4 Acoustic Leakage Analysis
2.4.1 Side-Channel Leakage Model
2.4.2 Leakage Quantification
2.4.3 Leakage Exploitation
2.5 Attack Model Description
2.5.1 Attack Model
2.5.2 Components of the Attack Model
2.5.2.1 Data Acquisition
2.5.2.2 Noise Filtering
2.5.2.3 Maximal Overlap Discrete Wavelet Transform and Multiresolution Analysis
2.5.2.4 Feature Extraction
2.5.2.5 Regression Model
2.5.2.6 Classification Model
2.5.2.7 Direction Prediction Model
2.5.2.8 Model Reconstruction
2.5.2.9 Post-Processing for Model Reconstruction
2.5.3 Attack Model Training and Evaluation
2.6 Results for Test Objects
2.6.1 Speed of Printing
2.6.2 The Dimension of the Object
2.6.3 The Complexity of the Object
2.6.4 Reconstruction of a Square
2.6.5 Reconstruction of a Triangle
2.6.6 Case Study: Outline of a Key
2.7 Discussion
2.7.1 Technology Variation
2.7.2 Sensor Position
2.7.3 Sensor Number
2.7.4 Dynamic Window
2.7.5 Feature Separation during Multiple Axis Movement and Noise
2.7.6 Target Machine Degradation
2.8 Summary
References
3 Aiding Data-Driven Attack Model with a Compiler Modification
3.1 Introduction
3.2 Attack Model Description
3.3 Compiler Attack
3.3.1 Profiling Phase
3.3.2 Attack Phase
3.3.3 Compiler Modification
3.3.4 Transformations for Leakage Maximization
3.4 Experimental Results
3.4.1 Accuracy Metric
3.4.2 Mutual Information
3.4.3 Partial Success Rate
3.4.4 Total Success Rate
3.5 Discussion
3.5.1 Countermeasures
3.6 Summary
References
Part II Data-Driven Defense of Cyber-Physical Systems
4 Data-Driven Defense Through Leakage Minimization
4.1 Introduction
4.1.1 Motivation for Leakage-Aware Security Tool
4.1.2 Problem and Challenges
4.1.3 Contributions
4.2 System Modeling
4.2.1 Data-driven Leakage Modeling and Quantification
4.2.2 Attack Model
4.2.3 Formulation of Data-Driven Leakage-Aware Optimization Problem
4.2.3.1 Design Variables for Leakage Minimization
4.2.3.2 Optimization Problem Statement
4.2.4 Success Rate of the Adversary
4.3 Experimental Results
4.3.1 Mutual Information
4.3.1.1 Design Variable—γ
4.3.1.2 Design Variable—v
4.3.2 Test with Benchmark 3D Models
4.4 Case Study with an Attack Model
4.4.1 Success Rate Calculation
4.4.2 Test Case with Reconstruction
4.5 Discussion
4.6 Summary
References
5 Data-Driven Kinetic Cyber-Attack Detection
5.1 Introduction
5.1.1 Motivation
5.1.2 Problem and Challenges
5.1.3 Contributions
5.2 Kinetic Cyber-Attack Adversary Model
5.3 KCAD Method
5.3.1 Mutual Information
5.3.2 KCAD Architecture
5.3.3 Acoustic Analog Emissions
5.3.4 Performance Metrics
5.4 Experimental Results
5.4.1 Experimental Setup
5.4.2 Mutual Information Calculation
5.4.3 Model Function Estimation
5.4.4 Results for Detection of Kinetic Attack
5.4.5 Test Case: Base Plate of a Quad Copter
5.5 Discussion
5.6 Summary
References
6 Data-Driven Security Analysis Using Generative Adversarial Networks
6.1 Introduction
6.1.1 Research Challenges
6.1.2 Preliminaries
6.1.3 Novel Contributions
6.2 CGAN-Based CPPS Security Model
6.3 CGAN Model Generation
6.4 Case Study and Analysis
6.4.1 GCPPS Generation
6.4.2 Experimental Data Collection
6.4.3 CGAN Modeling
6.4.4 Security Analysis Results
6.5 Summary
References
Part III Data-Driven Digital Twin Modeling
7 Dynamic Data-Driven Digital Twin Modeling
7.1 Introduction
7.1.1 Research Challenges
7.1.2 Contributions
7.1.3 Digital Twin Model
7.2 Digital Twin of Cyber-Physical Additive ManufacturingSystem
7.2.1 Key Performance Indicators (KPIs)
7.2.1.1 Surface Texture (K1)
7.2.1.2 Dimension (K2)
7.3 Keeping Digital Twin Updated
7.4 Building Digital Twin
7.4.1 Sensor/Emission Modality Selection
7.4.2 Feature Engineering
7.4.2.1 Time Domain
7.4.2.2 Frequency Domain
7.4.3 Sensor Positioning
7.4.4 Data-Driven Models
7.5 Experimental Setup
7.5.1 The Test-Bed
7.5.1.1 Sensors
7.5.1.2 Data Acquisition
7.5.1.3 Data Synchronization
7.5.2 Test 3D Objects
7.5.3 Data Collection
7.5.4 Data Segmentation
7.6 Simulation and Results for Digital Twin Models
7.6.1 Digital Twin Models
7.6.2 Aliveness
7.7 Summary
References
8 IoT-Enabled Living Digital Twin Modeling
8.1 Introduction
8.1.1 Research Challenges
8.1.2 Contribution
8.1.3 Motivational Case Study for Multi-Sensor DataAnalysis
8.1.4 Related Work
8.2 Background
8.2.1 Concept Definition
8.2.2 IoT Sensor Data as Side-Channels
8.2.3 Metric for Quality Measurement
8.3 Building the Digital Twin
8.3.1 DTproduct Parsing
8.3.2 Feature Extraction
8.3.3 Synchronize and Segment
8.3.4 Clustering Algorithm
8.3.5 Anomaly Localization Algorithm
8.3.6 Digital Twin Update Algorithm
8.3.7 Quality Inference Model
8.4 Experimental Setup
8.4.1 IoT Sensors
8.4.2 Digital Twin Parameters
8.4.3 Sensor Position Analysis
8.4.4 Performance of Clustering Algorithms
8.4.5 Anomaly Localization Accuracy
8.4.6 System Degradation Prediction Analysis
8.4.7 Quality Inference
8.4.8 Comparative Analysis
8.5 Discussion
8.6 Summary
References
Part IV Non-Euclidean Data-Driven Modeling of Cyber-Physical Systems
9 Non-euclidean Data-Driven Modeling Using Graph Convolutional Neural Networks
9.1 Introduction
9.2 Related Work
9.3 Graph Learning Using Convolutional Neural Network
9.3.1 Knowledge Graph Extraction
9.3.2 Attribute Embedding
9.3.3 Neighbor Nodes Aggregation
9.3.4 Structural Graph Convolutional Neural NetworkLayers
9.3.4.1 Sub-Graph Convolution Kernel
9.3.4.2 Graph Pooling Algorithm
9.3.4.3 2D Convolutions on Attribute Matrix
9.3.4.4 New Adjacency Matrix Calculation
9.3.5 Classification for Engineering Design Abstraction
9.3.6 Graph Learning Algorithm Hyper-Parameters
9.3.6.1 Path Length in Node Aggregation Layer
9.3.6.2 Graph Convolution Kernel Size
9.3.6.3 Dropout of Candidate Kernels
9.4 GrabCAD Dataset
9.5 Results
9.5.1 Activation Functions
9.5.2 Kernel Size
9.5.3 Dropout
9.5.4 Layers
9.6 Discussion
9.7 Summary
References
10 Dynamic Graph Embedding
10.1 Introduction
10.1.1 Research Challenges
10.1.2 Contribution
10.2 Related Work
10.2.1 Static Graph Embedding
10.2.2 Dynamic Graph Embedding
10.2.3 Dynamic Link Prediction
10.3 Motivating Example
10.4 Methodology
10.4.1 Problem Statement
10.4.2 dyngraph2vec Algorithm
10.4.3 Optimization
10.5 Experiments
10.5.1 Datasets
10.5.2 Baselines
10.5.3 Evaluation Metrics
10.6 Results and Analysis
10.6.1 SBM Dataset
10.6.2 Hep-th Dataset
10.6.3 AS Dataset
10.6.4 MAP Exploration
10.6.5 Hyper-Parameter Sensitivity: Lookback
10.6.6 Length of Training Sequence Versus MAP Value
10.7 Discussion
10.8 Summary
References
Index