StegaMultiPayload-2026: A Multi-Strategy and Multi-Payload Image Steganalysis Dataset

StegaMultiPayload-2026: A Multi-Strategy and Multi-Payload Image Steganalysis Dataset StegaMultiPayload-2026 is a large-scale benchmark dataset developed for image steganalysis research, supporting tasks such as steganography detection, embedding algorithm classification, and payload-length estimation. The dataset is designed to facilitate supervised learning and reproducible experimentation for machine learning and deep learning-based steganalysis systems. The dataset consists of 50,000 grayscale PNG images, including 25,000 original cover images and 25,000 corresponding stego images, organized in a structured and label-consistent hierarchy to support multi-task steganalysis. Dataset size is 5.75 GB. Dataset Highlights Total Images: 50,000 Cover Images: 25,000 Stego Images: 25,000 Image Format: PNG Image Type: Grayscale (Single Channel) Supported Resolutions: 512 × 512 and 256 × 256 Bit Depth: 8-bit Embedding Strategies Included The dataset includes stego images generated using five embedding strategy variants: LSB (Least Significant Bit) WOW-inspired HILL-inspired S-UNIWARD-inspired HUGO-inspired Each embedding strategy contains five payload categories: 5 words 10 words 15 words 20 words 25 words Each payload category contains 1,000 stego images, resulting in 5,000 images per embedding strategy and a balanced dataset structure for fair experimental evaluation. Repository Structure The repository includes: cover_data/ – Original cover images stego_data/ – Hierarchically organized stego images based on embedding strategy and payload length cover_dataset_index.csv – Metadata and indexing information for cover images stego_label_metadata.csv – Detailed annotation and labeling information for stego images Dataset_Description.pdf – Complete documentation of dataset structure, image properties, annotations, embedding strategies, metadata, and usage guidelines Research Applications StegaMultiPayload-2026 is suitable for research in: Image Steganalysis Steganography Detection Payload-Length Estimation Embedding Algorithm Classification Explainable AI for Steganalysis Machine Learning and Deep Learning-based Security Analytics Important Note The dataset employs inspired variants of classical embedding strategies rather than exact implementations, enabling controlled experimentation while maintaining diversity in embedding behavior. For complete dataset organization, annotation schema, file naming conventions, metadata specifications, embedding details, and usage instructions, refer to the accompanying Stego_Dataset_Description.pdf included in this repository. License: CC BY 4.0