Improved JPEG anti-forensics with better image visual quality and forensic undetectability

Area of Science:

• Digital Image Forensics
• Image Processing
• Computer Vision

Background:

• Digital image authenticity is crucial due to advanced manipulation tools.
• JPEG compression is commonly used in digital forensics for tampering detection.
• Existing forensic detectors require robust anti-forensic techniques for evaluation.

Purpose of the Study:

• To propose improved JPEG anti-forensic techniques.
• To remove blocking artifacts and reduce noise in JPEG compressed images.
• To enhance the trade-off between image visual quality and forensic undetectability.

Main Methods:

• Developed two improved JPEG anti-forensic techniques targeting spatial and DCT domains.
• Implemented denoising algorithms based on total variation energy minimization and normalized weighted functions.
• Applied a TV-based deblocking operation and a decalibration operation.

Main Results:

• Proposed techniques effectively reduce blocking artifacts and grainy noise.
• Achieved superior performance compared to state-of-the-art methods.
• Demonstrated an enhanced trade-off between image visual quality and forensic undetectability.

Conclusions:

• The proposed anti-forensic approaches offer significant improvements in JPEG image forensics.
• These techniques provide a valuable tool for evaluating the robustness of forensic detectors.
• High computational cost remains a consideration for practical implementation.