Robustness

Description

This dataset contains the evaluation files for robustness testing against visual manipulations and geometric attacks. It includes the original baseline image (Stego_Original.jpg), seven modified images subjected to various simulated attacks (e.g., Gaussian Noise, Blur, Cropping, JPEG Compression, Scaling, Rotation, and Brightness adjustments), and a structured dataset (Robustness_Test_Results_2.csv). The CSV file provides the precise quantitative evaluation metrics for each scenario, specifically measuring visual fidelity and extraction reliability through Mean Squared Error (MSE), Peak Signal-to-Noise Ratio (PSNR), Structural Similarity Index (SSIM), and Bit Error Rate (BER).

Steps to reproduce

1. Prepare the Baseline: Utilize the pristine digital stego-image (Stego_Original.jpg) as the primary reference document. 2. Apply Manipulations: Subject the baseline image to distinct visual manipulations and geometric transformations, including 10% cropping, 50% scaling, 5-degree rotation, Gaussian noise addition, Gaussian blur, brightness adjustments, and JPEG-80 compression. 3. Calculate Visual Metrics: Compute the structural degradation and visual fidelity of the manipulated images by comparin3. g them against the baseline to obtain the MSE, PSNR, and SSIM values. 4. Evaluate Payload Extraction: Execute the data extraction mechanism on each manipulated image to determine the data loss or extraction sensitivity, recorded as the Bit Error Rate (BER). 5. Aggregate Results: Compile all calculated metrics corresponding to their respective attack scenarios into the final structured report

Institutions

• Universitas Ahmad Dahlan

  Yogyakarta, Yogyakarta

Categories

Licence