Comparative Analysis of Steganography Injection and Detection Tools/Methods: Traditional vs. AI-Based

Document Type

Conference Proceeding

Source of Publication

Proceedings of the International Symposium on Digital Forensics and Security Isdfs

Publication Date

3-19-2026

Abstract

The paper presents a cross-media benchmarking study of steganography injection and steganalysis tools for images, audio, and video under a unified protocol. Traditional and learning-based methods are compared using shared datasets, fixed payload settings, and standardized metrics, and a weighted scoring rubric is introduced to summarize imperceptibility, robustness, security, computational cost, and detection behaviour. Results show that highly imperceptible image embedding (e.g., LSB-based tools with ≈ 0.01% pixel modification) can still be detected reliably, while aggressive embedding (e.g., F5 with 98.78% pixel modification) produces visible artifacts and remains detectable. For audio and video, metadata, and structure-based indicators enable reliable detection, with the combined audio detection strategy achieving 100% identification of stego files while preserving clean-file accuracy. These results support evidence-based tool selection for digital forensics triage and secure deployment under operational constraints.

ISBN

[9798331573102]

ISSN

2768-1831

Publisher

IEEE

Issue

2026

Disciplines

Computer Sciences

Keywords

Chi-Square Test, Deep Learning, Digital Forensics, Discrete Cosine Transform (DCT), Least Significant Bit (LSB), Machine Learning, RS Analysis, Steganalysis, Steganography

Scopus ID

105038359979

Indexed in Scopus

yes

Open Access

no

Share

COinS