Machine Learning and Cybersecurity

  1. Adversarial Machine Learning and Security
    1. Threat Model for ML Systems
      1. Adversary Goals
        1. Integrity Attacks
          1. Availability Attacks
            1. Privacy Attacks
            2. Adversary Knowledge
              1. Perfect Knowledge (White-box)
                1. Limited Knowledge (Black-box)
                  1. Partial Knowledge (Gray-box)
                  2. Adversary Capabilities
                    1. Training Data Manipulation
                      1. Test Data Manipulation
                        1. Model Access
                      2. Types of Adversarial Attacks
                        1. Evasion Attacks
                          1. Adversarial Examples
                            1. Fast Gradient Sign Method (FGSM)
                              1. Projected Gradient Descent (PGD)
                                1. Carlini & Wagner (C&W) Attack
                                2. Feature Manipulation
                                  1. Gradient-Based Methods
                                    1. Optimization-Based Methods
                                    2. Domain-Specific Evasion
                                      1. Malware Evasion
                                        1. Network Traffic Evasion
                                          1. Spam Evasion
                                        2. Poisoning Attacks
                                          1. Training Data Poisoning
                                            1. Label Flipping
                                              1. Data Injection
                                              2. Backdoor Attacks
                                                1. Trigger-Based Backdoors
                                                  1. Clean-Label Backdoors
                                                  2. Model Poisoning
                                                    1. Federated Learning Poisoning
                                                  3. Extraction Attacks
                                                    1. Model Stealing
                                                      1. Equation-Solving Attacks
                                                        1. Path-Finding Attacks
                                                        2. Property Inference
                                                          1. Membership Inference
                                                            1. Shadow Model Attacks
                                                              1. Threshold Attacks
                                                            2. Inversion Attacks
                                                              1. Model Inversion
                                                                1. Property Inference
                                                                  1. Attribute Inference
                                                                2. Defenses Against Adversarial Attacks
                                                                  1. Proactive Defenses
                                                                    1. Adversarial Training
                                                                      1. Min-Max Optimization
                                                                        1. Adversarial Example Generation
                                                                        2. Defensive Distillation
                                                                          1. Temperature Scaling
                                                                            1. Knowledge Transfer
                                                                            2. Certified Defenses
                                                                              1. Randomized Smoothing
                                                                                1. Interval Bound Propagation
                                                                                2. Robust Optimization
                                                                                  1. Distributionally Robust Optimization
                                                                                    1. Worst-Case Training
                                                                                  2. Reactive Defenses
                                                                                    1. Input Preprocessing
                                                                                      1. Feature Squeezing
                                                                                        1. Spatial Smoothing
                                                                                          1. JPEG Compression
                                                                                          2. Adversarial Detection
                                                                                            1. Statistical Tests
                                                                                              1. Neural Network Detectors
                                                                                                1. Reconstruction-Based Detection
                                                                                                2. Input Transformation
                                                                                                  1. Random Transformations
                                                                                                    1. Defensive Transformations
                                                                                                  2. Architectural Defenses
                                                                                                    1. Ensemble Methods
                                                                                                      1. Diverse Model Ensembles
                                                                                                        1. Randomized Ensembles
                                                                                                        2. Network Architecture Modifications
                                                                                                          1. Defensive Layers
                                                                                                            1. Gradient Masking
                                                                                                        3. Evaluation of Adversarial Robustness
                                                                                                          1. Robustness Metrics
                                                                                                            1. Adversarial Accuracy
                                                                                                              1. Certified Accuracy
                                                                                                                1. Attack Success Rate
                                                                                                                2. Evaluation Methodologies
                                                                                                                  1. Adaptive Attacks
                                                                                                                    1. Gradient-Free Attacks
                                                                                                                      1. Transfer Attacks
                                                                                                                      2. Benchmarking
                                                                                                                        1. Standard Datasets
                                                                                                                          1. Evaluation Frameworks
                                                                                                                            1. Robustness Competitions

                                                                                                                        Previous

                                                                                                                        4. Advanced Machine Learning Techniques

                                                                                                                        Go to top

                                                                                                                        Next

                                                                                                                        6. Model Development and Deployment