Comprehensive Dataset for Analyzing Detectability in LSB Steganography

Öztürk E., Mesut A., Aydın Fidan Ö.

INTERNATIONAL JOURNAL OF INFORMATICS TECHNOLOGIES, cilt.19, sa.1, ss.79-90, 2026 (TRDizin)

Özet

Steganography involves concealing secret information within selected cover media to prevent detection of its existence, complementing cryptography by enhancing information security. Among various steganographic methods, Least Significant Bit (LSB) embedding stands out due to its simplicity, efficiency, and widespread applicability, particularly in digital images. Although many researchers have studied steganographic methods, there is still a lack of standard datasets to effectively test and compare LSB methods. To address this gap, this study introduces a large and well-organized dataset consisting of 432,000 steganographic images. The images are selected from high-resolution sources and resized into four different sizes: 128×128, 256×256, 512×512, and 1024×1024 pixels. Secret messages hidden within these images are drawn from texts in six languages—German, English, Spanish, French, Italian, and Turkish—to analyze linguistic effects on steganographic detectability. Each image includes detailed metadata about the hidden information, such as the number of bits used per pixel, image size, message length, and whether the embedding is sequential or random. The dataset enables researchers to investigate how factors like message length, embedding randomness, and language differences impact the detectability of hidden messages. It provides a valuable resource for developing and evaluating new deep learning techniques in steganography and enhancing digital information security methods.