Digital Forensics and Watermarking: 20th International Workshop, IWDW 2021, Beijing, China, November 20–22, 2021, Revised Selected Papers

Description

This volume constitutes the proceedings of the 20th International Workshop on Digital Forensics and Watermarking, IWDW 2021, held in Beijing, China, in November 2021. The 18 full papers in this volume were carefully reviewed and selected from 32 submissions. The are categorized in the following topical headings: Forensics and Security Analysis; Watermarking and Steganology. Preface Organization Contents Forensics and Security Analysis A Multi-level Feature Enhancement Network for Image Splicing Localization 1 Introduction 2 Preliminary 3 Proposed Method 3.1 Manipulation Detection 3.2 Splicing Segmentation 3.3 Fused Loss Function 4 Experiments 4.1 Preparation 4.2 Evaluation 5 Conclusion References MSA-CNN: Face Morphing Detection via a Multiple Scales Attention Convolutional Neural Network 1 Introduction 2 Related Work 2.1 Vulnerability of FRS on Face Morphing Attack 2.2 Face Morphing Detection 3 The Proposed MSA-CNN 3.1 Overview of MSA-CNN 3.2 Design of Attention Convolutional Neural Network 3.3 Design of Morphing Artifacts Indication Network 3.4 Design of Multi-scale Classifier and Multi-scale Classification Loss 4 Experiments and Analysis 4.1 Database and Evaluation Metrics 4.2 Implementation Details 4.3 Experiment Results 5 Conclusion References Double-Stream Segmentation Network with Temporal Self-attention for Deepfake Video Detection 1 Introduction 2 Related Works 3 Proposed Method 3.1 Double-Stream Fully Convolutional Network 3.2 Temporal Double-Stream Fully Convolutional Network 4 Experiments 4.1 Experimental Settings 4.2 Experimental Results 4.3 Ablation Study 5 Conclusion References Exposing Deepfake Videos with Spatial, Frequency and Multi-scale Temporal Artifacts 1 Introduction 2 Proposed Method 2.1 Overall Framework 2.2 DCT Feature Extraction 2.3 Short-Term Temporal Feature Extraction 2.4 Long-Term Temporal Feature Extraction 3 Experiments 3.1 Settings 3.2 Ablation Study 3.3 Comparison with Related Works 4 Conclusion References On the Security of Encrypted JPEG Image with Adaptive Key Generated by Invariant Characteristic 1 Introduction 2 The Basis of CPA for Hes ACCs Encryption 2.1 Basis of Same Adaptive Encryption Key 2.2 Basis of Attacking MCUs Permutation 2.3 Basis of Attacking ACCs Permutation with Same Run-Length 3 The Proposed CPA 3.1 Plaintext Image Construction 3.2 Image Encryption 3.3 Attacking MCUs Permutation 2 3.4 Attacking ACCs Permutation 1 3.5 ACCs Decrypted Image Generation 4 Experimental Result 4.1 Analysis of Attacking 2 4.2 Analysis of Attacking 1 5 Conclusion References Modify the Quantization Table in the JPEG Header File for Forensics and Anti-forensics 1 Introduction 2 The Role of Quantization Table in JPEG Decompression Process 3 The Proposed Method 3.1 Anti-forensics Model 3.2 Forensics Model 4 Experimental Results 4.1 The Performance of Anti-forensics Model 4.2 The Performance of Forensics Model 5 Conclusion References More Accurate and Robust PRNU-Based Source Camera Identification with 3-Step 3-Class Approach 1 Introduction 2 Review of Previous Works 2.1 PRNU-Based Algorithm and Source Camera Identification 2.2 Fingerprint-Copy Attack 2.3 Local Image Alteration Attack 3 Proposed Method 3.1 3-Class Classification 3.2 Excess Correlation Analysis Against Fingerprint-Copy Attacks 3.3 Hierarchical Multi-window Search for Forgery Localization 4 Experimental Setting 5 Experimental Results 5.1 Source Camera Identification Accuracy 5.2 Fingerprint-Copy Attack Detection Performance 5.3 Forgery Localization Performance 6 Conclusions References Effects of Image Compression on Image Age Approximation 1 Introduction 2 In-Field Sensor Defects 3 Image Age Approximation Methods 4 Defect Detection Methods 5 Experiments 6 Experimental Results 7 Conclusion References FMFCC-A: A Challenging Mandarin Dataset for Synthetic Speech Detection 1 Introduction 2 Dataset and Baselines 2.1 FMFCC-A Dataset and Partitions 2.2 Baseline Methods 3 Evaluation Metrics and FMFCC-A Schedule 3.1 Evaluation Metrics 3.2 FMFCC-A Schedule 4 FMFCC-A Results and Discussion 4.1 Preliminary Phase of FMFCC-A 4.2 Final Evaluation Phase of FMFCC-A 4.3 Performance on Unknown Speech Synthesis Systems 4.4 Performance on Audio Post-processing Operations 5 Conclusion References Watermarking A Feature-Map-Based Large-Payload DNN Watermarking Algorithm 1 Introduction 2 State of the Art 2.1 Static Watermarking 2.2 Dynamic Watermarking 3 The Proposed Dynamic Watermarking Algorithm 3.1 Requirements of DNN Watermarking 3.2 The Proposed Feature-Map-Based Dynamic Watermarking Algorithm 3.3 Choice of the Key Images 4 Experimental Results 4.1 Datasets and Host Networks 4.2 Payload, Bit Error Rate and Test Error Rate 4.3 Robustness Evaluation 5 Conclusion References A Robust DCT-Based Video Watermarking Scheme Against Recompression and Synchronization Attacks 1 Introduction 2 Proposed Method 2.1 Pre-processing 2.2 Watermark Embedding 2.3 Watermark Extraction 3 Experimental Results 4 Conclusions References Improved Fluctuation Derived Block Selection Strategy in Pixel Value Ordering Based Reversible Data Hiding 1 Introduction 2 Related Work 3 The Proposed Scheme 3.1 Framework 3.2 Calculation of Fluctuation Value 3.3 Secret Data Embedding 3.4 Secret Data Extraction 4 Experimental Results and Analysis 4.1 Verification of Reversibility 4.2 Comparison of Imperceptibility 5 Conclusion References MasterFace Watermarking for IPR Protection of Siamese Network for Face Verification 1 Introduction 2 Background on DNN Watermarking 3 The Proposed MF Watermarking Algorithm 3.1 Face Verification (FV) System 3.2 Watermarking Model and Requirements 3.3 MF Watermarking Algorithm 4 Experimental Methodology and Setting 4.1 Evaluation Metrics 4.2 Datasets 4.3 Other Datasets 4.4 Network Implementation and Settings 5 Results 5.1 Performance Analysis 5.2 Robustness Analysis 6 Conclusion References Steganology Arbitrary-Sized JPEG Steganalysis Based on Fully Convolutional Network 1 Introduction 2 Related Works 3 The Proposed Method 3.1 Architecture 3.2 Detailed Design of EWNet 4 Setup of Experiments 4.1 Datasets 4.2 Implementation Details 5 Experimental Results and Analysis 5.1 Evaluation on Fixed-Size Dataset 5.2 Evaluation on Arbitrary-Sized Dataset 6 Conclusion References Using Contrastive Learning to Improve the Performance of Steganalysis Schemes 1 Introduction 2 Background 2.1 DNN-Based Steganalysis 2.2 The Loss of Contrastive Learning 3 Proposed Methods 3.1 Steganalysis Contrastive Framework 3.2 Steganalysis Contrastive Loss 4 Experiments 4.1 Dataset 4.2 Metric 4.3 Detection Accuracy 4.4 Generalization 4.5 Computing Complexity 5 Conclusion References A HEVC Steganalysis Algorithm Based on Relationship of Adjacent Intra Prediction Modes 1 Introduction 2 Modelling of IPM-Based HEVC Steganography 2.1 Basics of Intra Prediction in HEVC 2.2 Modelling of IPM-Based HEVC Steganography 3 Proposed RoAIPM Feature 4 Proposed Steganalysis Algorithm 5 Experiments 5.1 Experimental Setup 5.2 Performance on Accuracy Rate 5.3 Performance with Dimension Reduction 5.4 Performance on Feature Extraction Time 6 Conclusion References Data Hiding Based on Redundant Space of WeChat Mini Program Codes 1 Introduction 2 Preliminaries 3 The Proposed Scheme 3.1 Overall Design 3.2 Embedding Process 3.3 Extraction Process 3.4 Theoretical Analysis 4 Experiment and Evaluation 4.1 Experiments 4.2 Subjective Evaluation 4.3 Comparison 5 Conclusion References Image Block Regression Based on Feature Fusion for CNN-Based Spatial Steganalysis 1 Introduction 2 Proposed Method 2.1 Backbone Network 2.2 Block Regression Module 2.3 Loss Function 3 Experiments 3.1 Datasets 3.2 Training Settings 3.3 Evaluation Metric 3.4 Results 3.5 Time Performance Comparison 4 Conclusion References Author Index