(Downloads - 0)
For more info about our services contact : help@bestpfe.com
Table of contents
Acknowledgements
I Context and State of the Art
1 Context, Objectives and Contributions
1.1 Introduction to Cryptography
1.1.1 Description of AES
1.2 Secure Components
1.2.1 Embedded Cryptography Vulnerabilities
1.2.1.1 Side-Channel Attacks
1.2.1.2 A Classification of the Attacks against Secure Components
1.2.2 Certification of a Secure Hardware – The Common Criteria
1.2.2.1 The actors
1.2.2.2 The Target of Evaluation and the security objectives
1.2.2.3 Evaluation Assurance Level and Security Assurance Requirements
1.2.2.4 The AVA_VAN family and the Attack Potential
1.2.2.5 The Evaluation Technical Report
1.3 This thesis objectives and contributions
1.3.1 The Preliminary Purpose of this Thesis: Research of Points of Interest
1.3.2 Dimensionality Reduction Approach
1.3.3 Towards Machine Learning and Neural Networks Approach
2 Introduction to Side-Channel Attacks
2.1 Notations and Probability and Statistics Recalls
2.2 Side-Channel Attacks: an Overview
2.3 Physical Nature of the Exploited Signals
2.4 Sensitive Variables
2.5 The Strategy Family
2.5.1 Simple Attacks
2.5.2 Collision Attacks
2.5.3 Advanced Attacks
2.6 The Shape of the Attack
2.7 The Attacker Knowledge
2.8 Efficiency of the SCAs
2.9 Advanced Attacks
2.9.1 Leakage Models
2.9.2 Distinguishers
2.10 Profiling Side-Channel Attacks
2.10.1 Template Attack
2.10.1.1 The Curse of Dimensionality
2.10.1.2 The Gaussian Hypothesis
2.10.2 Points of Interest and Dimensionality Reduction
2.11 Main Side-Channel Countermeasures
2.11.1 Hiding
2.11.2 Masking
3 Introduction to Machine Learning
3.1 Basic Concepts of Machine Learning
3.1.1 The Task, the Performance and the Experience
3.1.2 Example of Linear Regression
3.1.3 Example of Linear Model for Classification
3.1.4 Underfitting, Overfitting, Capacity, and Regularization
3.1.5 Hyper-Parameters and Validation
3.1.6 No Free Lunch Theorem
3.2 Overview of Machine Learning in Side-Channel Context
II Contributions
4 Linear Dimensionality Reduction
4.1 Introduction
4.2 Principal Component Analysis
4.2.1 Principles and algorithm description
4.2.2 Original vs Class-Oriented PCA
4.2.3 Computational Consideration
4.2.4 The Choice of the Principal Components
4.2.4.1 Explained Local Variance Selection Method
4.3 Linear Discriminant Analysis
4.3.1 Fisher’s Linear Discriminant and Terminology Remark
4.3.2 Description
4.3.3 The Small Sample Size Problem
4.3.3.1 Fisherface Method
4.3.3.2 SW Null Space Method
4.3.3.3 Direct LDA
4.3.3.4 ST Spanned Space Method
4.4 Experimental Results
4.4.1 The testing adversary
4.4.2 Scenario 1
4.4.3 Scenario 2
4.4.4 Scenario 3
4.4.5 Scenario 4
4.4.6 Overview of this Study and Conclusions
4.5 Misaligning Effects
5 Kernel Discriminant Analysis
5.1 Motivation
5.1.1 Getting information from masked implementations
5.1.2 Some strategies to perform higher-order attacks
5.1.2.1 Higher-Order Version of Projection Pursuits
5.1.3 Purpose of this Study
5.2 Feature Space, Kernel Function and Kernel Trick
5.3 Kernel Discriminant Analysis
5.3.1 KDA for dth-order masked side-channel traces
5.3.2 The implicit coefficients
5.3.3 Computational complexity analysis
5.4 Experiments over Atmega328P
5.4.1 Experimental Setup
5.4.2 The Regularisation Problem
5.4.3 The Multi-Class Trade-Off
5.4.4 Asymmetric Preprocessing/Attack Approach
5.4.5 Comparison with Projection Pursuits
5.5 Conclusions and Drawbacks
6 Convolutional Neural Networks
6.1 Motivation
6.2 Introduction
6.3 Neural Networks and Multi-Layer Perceptrons
6.4 Learning Algorithm
6.4.1 Training
6.4.2 Cross-Entropy
6.5 Attack Strategy with an MLP
6.6 Performance Estimation
6.6.1 Maximal Accuracies and Confusion Matrix
6.6.2 Side-Channel-Oriented Metrics
6.7 Convolutional Neural Networks
6.8 Data Augmentation
6.9 Experiments against Software Countermeasures
6.9.1 One Leaking Operation
6.9.2 Two Leaking Operations
6.10 Experiments against Artificial Hardware Countermeasures
6.10.1 Performances over Artificial Augmented Clock Jitter
6.11 Experiments against Real-Case Hardware Countermeasures
6.12 Conclusion
7 Conclusions and Perspectives
7.1 Conclusions
7.2 Tracks for FutureWorks
A Cross-Validation
B Artificially Simulated Jitter
C Kernel PCA construction
C.1 Kernel class-oriented PCA
Bibliography
