AI Embedded Assurance for Cyber Systems

• Autore: Wang Cliff
• Pubblicazione: Cham : , : Springer International Publishing AG, , 2023; ©2023
• Edizione: 1st ed.
• Descrizione fisica: 1 online resource (252 pages)
• Disciplina: 006.3
• Altri autori: IyengarS. S SunKun
• Nota di contenuto: Intro -- Foreword -- Preface -- Acknowledgements -- Contents -- Contributors -- Acronyms -- Part I AI/ML for Digital Forensics -- 1 Writer-Dependent Off-Line Signature Verification with Neural Networks -- 1.1 Introduction -- 1.2 A Description of the Verification System -- 1.3 Preprocessing -- 1.3.1 Loading the Image and Conversion to Grayscale -- 1.3.2 Cropping the Image -- 1.3.3 Binarization -- 1.3.4 Resizing the Image -- 1.3.5 Thinning (Skeletonization) -- 1.4 Feature Extraction -- 1.4.1 Global Features -- 1.4.1.1 Number of Signature Pixels (in Thinned Image) -- 1.4.1.2 Area of the Signature (in Resized Image) -- 1.4.1.3 Aspect Ratio (in Binary Image) -- 1.4.1.4 The Inclination Angle of the Line Joining the Center of Gravity and the Lower Right Corner (in Binary Image) -- 1.4.1.5 Vertical Variance and Horizontal Variance (in Binary Image) -- 1.4.1.6 Intersection Points and Border Points (in Thinned Image) -- 1.4.2 Local Features -- 1.4.2.1 Pixel Density -- 1.4.2.2 Pixel Angle -- 1.4.2.3 Pixel Distance -- 1.5 Pairing and Classification -- 1.5.1 Pairing -- 1.5.2 Classification -- 1.5.2.1 NN Architecture -- 1.5.2.2 Training NN -- 1.6 Verification Process -- 1.7 Experiment Results -- 1.7.1 ICDAR 2011 SigComp Database -- 1.7.2 GPDS Synthetic Signature Corpus -- 1.7.3 Experiment Results -- 1.7.3.1 Evaluation Metrics -- 1.7.3.2 ICDAR 2011 Database Test Results -- 1.7.3.3 GPDS Corpus Test Results -- 1.8 Summary and Future Works -- References -- 2 Political Activism and Technology -- 2.1 Activism and Technology Use -- 2.2 Threat Models and the Technical-Defensive Landscape -- 2.2.1 Shaping Threat Models Through the Technical Capabilities of Political Allies and Enemies -- 2.2.2 The Power of the State to Compel Authentication -- 2.2.3 Control over the Telecommunication Infrastructure -- 2.3 Societal Context and Technology Adoption. ; 2.3.1 Institutional Knowledge Sharing-Security and Privacy Advice -- 2.3.2 Building Trust in a Mutating Group Surrounded by Uncertainty -- 2.3.3 Support From Abroad -- 2.4 Conclusions-Needs and Technology -- References -- 3 Forensic Proof and Criminal Liability for Development, Distribution and Use of Artificial Intelligence -- 3.1 Artificial Intelligence and Criminal Liability -- 3.2 Injuries as Crimes, from Loss of Life to Loss of Liberties, and Policies of Review -- 3.3 Investigation, AI Forensics and Proof of Responsibility -- 3.4 The Ethics of Artificial Intelligence -- 3.5 Conclusion -- References -- Part II AI/ML for CPS -- 4 Automotive Batteries as Anomaly Detectors -- 4.1 Introduction -- 4.2 Prototpe and Data Collection -- 4.3 Case-Study: Detecting Engine Anomalies Using Batteries -- 4.3.1 Automotive Battery and Vehicle Engine -- 4.3.2 Detecting RPM Anomalies with Battery -- 4.3.2.1 Data Preparation -- 4.3.2.2 Norm Model Construction -- 4.3.2.3 Anomaly Detection -- 4.3.2.4 Anomaly Verification -- 4.4 Detecting Vehicle Anomaliues Beyond Enginen RPM -- 4.5 Evaluations -- 4.5.1 B-Diag Against ``True'' Anomalies -- 4.5.1.1 Methodology -- 4.5.1.2 Evaluation Results -- 4.5.1.3 Adapter Faults or Vehicle Faults? -- 4.5.2 B-Diag Against Emulated Anomalies -- 4.5.2.1 Anomaly Model -- 4.5.2.2 Evaluation with Subaru Crosstrek -- 4.5.2.3 Evaluation with Other Vehicles -- 4.5.2.4 Diagnosing Beyond Engine RPM -- 4.6 Conclusions -- References -- 5 Zero Trust Architecture For Cyber-Physical Power System Security Based on Machine Learning -- 5.1 Introduction -- 5.2 Overview of Cyber-Physical Power System Security -- 5.2.1 The Hierarchical Structure for Cyber-Physical Power System -- 5.2.2 Cyber-Physical Power System Security -- 5.2.3 Examples for Cross-layer Failures in CPPS -- 5.3 Machine Learning Application in Cyber-Physical Power System Security. ; 5.3.1 Challenges in Model-Based Approaches for CPPS Security -- 5.3.2 Machine Learning Approaches for CPPS Security -- 5.4 A Combination of Novel Security Technique and Machine Learning-Based Approaches -- 5.4.1 Zero Trust Architecture Basics -- 5.4.1.1 Variations of Zero Trust Architecture Techniques -- 5.4.1.2 Related Work and Research Gap -- 5.4.2 Zero Trust Architecture Dedicated to Cyber-Physical Power System -- 5.4.3 Dynamic Trust Evaluation in Score-Based Policy Engine -- 5.4.3.1 Measurement Recovery in State Estimation -- 5.4.3.2 Shedding Loads in Frequency Recovery -- 5.4.4 How Machine Learning Empowers Zero Trust Architecture -- 5.4.4.1 User and Entity Behavioral Analytics Based on ML -- 5.4.4.2 Dynamic Access Control with at Least Privilege -- 5.5 Conclusion -- References -- 6 AI-enabled Real-Time Sensor Attack Detection for Cyber-Physical Systems -- 6.1 Introduction -- 6.2 Background -- 6.2.1 Cyber-Physical Systems -- 6.2.2 Sensor Attacks in CPSs -- 6.3 AI-enabled Sensor Attack Detection -- 6.3.1 Sensor Attack Detection Tasks -- 6.3.2 Workflow for AI-based Detector Design -- 6.3.3 Neural Network Model Selection -- 6.3.4 Case Studies -- 6.4 Detection Exploring Inherent Sensor Redundancy -- 6.4.1 Background of Auto-Encoder -- 6.4.2 Methodology -- 6.4.2.1 How to Forward Sensor Data to Autoencoder -- 6.4.2.2 Reconstruction Error Measurement -- 6.4.2.3 Threshold Estimation -- 6.4.3 Results -- 6.4.3.1 Experimental Setup -- 6.4.3.2 Performance Analysis -- 6.5 Real-Time Adaptive Sensor Attack Detection -- 6.5.1 System Design -- 6.5.1.1 Attack Detector -- 6.5.1.2 Behavior Predictor -- 6.5.1.3 Design of Drift Adaptor -- 6.5.2 Results -- 6.6 Conclusion -- References -- Part III AI/ML for Cyber Analysis -- 7 Generating Vulnerable Code via Learning-Based Program Transformations -- 7.1 Introduction -- 7.2 Mining Existing Vulnerability Data. ; 7.3 Learning-Based Data Generation via Vulnerability Injection -- 7.3.1 Problem Formulation -- 7.3.2 Model Design -- 7.3.3 Model Training and Testing -- 7.4 Technique Implementation -- 7.5 Empirical Evaluation of Performance -- 7.5.1 Experimental Setup -- 7.5.2 Evaluation Results -- 7.6 Discussion -- 7.6.1 Data Characteristics -- 7.6.2 Technical Limitations -- 7.7 Conclusion -- References -- 8 10 Security and Privacy Problems in Large Foundation Models -- 8.1 Introduction -- 8.2 Background on Self-Supervised Learning -- 8.2.1 Self-Supervised Learning in NLP -- 8.2.1.1 Pre-Training a Language Model -- 8.2.1.2 Fine-Tuning a Language Model for a Downstream Task -- 8.2.2 Self-Supervised Learning in CV -- 8.2.2.1 Pre-Training an Image Encoder (and a Text Encoder) -- 8.2.2.2 Applying an Image Encoder (and a Text Encoder) to Downstream Tasks -- 8.2.3 Self-Supervised Learning in Graph -- 8.2.3.1 Pre-Training a Graph Encoder -- 8.2.3.2 Applying a Graph Encoder to Downstream Tasks -- 8.3 Six Problems on Confidentiality -- 8.4 Three Problems on Integrity -- 8.5 One Problem on Availability -- 8.6 Conclusion -- References -- 9 Federated Learning for IoT Applications, Attacks and Defense Methods -- 9.1 Introduction -- 9.2 Background of Federated Learning -- 9.3 Internet of Things FL -- 9.3.1 General Applications of IoT FL -- 9.3.2 Secure Applications of IoT FL -- 9.4 Threat Models in FL -- 9.4.1 Outside Attacks in FL -- 9.4.2 Inside Attacks in FL -- 9.5 Defense Methods Against Attacks in FL -- 9.5.1 Defense Methods Against Adversarial Attacks in FL -- 9.5.2 Defense Methods Against Privacy Attacks in FL -- 9.6 Conclusion -- References -- 10 AI Powered Correlation Technique to Detect Virtual Machine Attacks in Private Cloud Environment -- 10.1 Introduction -- 10.2 Related Works -- 10.3 CORRNET: Correlational Neural Network -- 10.3.1 CorrNet Implementation. ; 10.3.2 Analysis of CorNet -- 10.3.3 Classification of Data -- 10.3.4 Detection -- 10.3.5 Correlation -- 10.4 Threat Prediction and Protection Algorithm -- 10.4.1 Technique Customization -- 10.4.2 Implementation of Algorithmic Structure -- 10.5 Case Study: Digital Payment Service -- 10.6 Conclusion -- References -- 11 Detecting Fake Users in Online Social Networks -- 11.1 Introduction -- 11.2 Related Work -- 11.2.1 The Behavior-Based Sybil Attack Detection -- 11.2.2 The Structure-Based Sybil Attack Detection -- 11.3 Detect Sybil Accounts at Registration and Growing-Up Stages -- 11.3.1 Overview -- 11.3.2 Detecting Sybil Accounts Using the Registration Information -- 11.3.2.1 Feature Extraction -- 11.3.2.2 Registration Graph Building -- 11.3.2.3 Sybil Accounts Detection -- 11.3.3 Detecting Sybil Accounts with Growing-Up Behaviors -- 11.3.3.1 Account-Behavior Bigraph Construction -- 11.3.3.2 Account-Account Graph Construction -- 11.3.3.3 Unsupervised Maliciousness Assessment -- 11.3.3.4 Growing-Up Sybil Account Detection -- 11.4 New Challenges in Defending Sybil Attack -- 11.5 Conclusion -- References -- 12 Explaining Deep Learning Based Security Applications -- 12.1 Introduction -- 12.2 Explainable Machine Learning -- 12.2.1 Problem Definition -- 12.2.2 Whitebox Explanation Methods -- 12.2.3 Blackbox Explanation Methods -- 12.3 Explaining Security Applications -- 12.3.1 Deep Learning in Security Applications -- 12.3.2 Why Not Existing Explanation Methods -- 12.4 Our Explanation Method -- 12.4.1 Insights Behind Our Designs -- 12.4.2 Model Development -- 12.4.3 Applying the Model for Explanation -- 12.5 Evaluation -- 12.5.1 Experimental Setup -- 12.5.2 Fidelity Evaluation -- 12.5.3 Experimental Results -- 12.6 Applications of ML Explanation -- 12.6.1 Understanding Classifier Behavior -- 12.6.2 Troubleshooting Classification Errors. ; 12.6.3 Targeted Patching of ML Classifiers.
• Titolo autorizzato: AI Embedded Assurance for Cyber Systems
• ISBN: 3-031-42637-1
• Formato: Materiale a stampa
• Livello bibliografico: Monografia
• Lingua di pubblicazione: Inglese
• Record Nr.: 9910770267403321
• Lo trovi qui: Univ. Federico II