Does this extract the hidden message?

No. It is a detector, not an extractor. It uses statistics to estimate whether sequential LSB embedding is likely present and reports a likelihood, because extraction needs the embedding scheme, ordering and any password.

Why only PNG and BMP, not JPEG?

LSB steganography hides data in the lowest bit of pixel values, and JPEG's lossy compression rewrites those values, destroying the payload. Only lossless formats like PNG and BMP preserve LSBs, so detection is meaningful there.

How does the chi-squared test work?

Natural images have uneven counts within each pair of values that differ only in the last bit (e.g. 100 vs 101). Sequential LSB embedding flattens those pairs toward equality. The chi-squared statistic measures that flattening across all value pairs; a large value indicates likely embedding.

Can it give false positives?

Yes. Heavily processed, dithered or synthetic images can naturally have balanced LSBs and may score high, while data hidden in only part of an image may score low overall. Treat the result as a signal to investigate, not proof.

Is my image uploaded?

No. The image is decoded into a Canvas in your browser and all pixel math runs in JavaScript locally. Nothing is sent to a server, which matters for sensitive forensic material.

What is the Image Steganography Detector?

Free image steganography detector. Loads a PNG or BMP and runs a chi-squared statistical test on pixel least-significant bits via Canvas to estimate the likelihood of hidden LSB-embedded data. It reports a likelihood, not the payload. Runs entirely in your browser. It runs free in your browser on Gera Tools, with nothing uploaded.

Image Steganography Detector

Name: Image Steganography Detector
Creator: Gera Tools
License: https://creativecommons.org/licenses/by/4.0/

Estimate whether an image hides data

LSB (least-significant-bit) steganography hides a message by overwriting the lowest bit of each colour value — a change invisible to the eye. This tool loads a lossless image (PNG or BMP), reads every pixel via the Canvas API, and applies statistical steganalysis to estimate whether such data is present. It reports a likelihood; it does not extract the payload, because that requires knowing the exact embedding scheme, ordering, and any encryption key.

How it works

The image is drawn to an off-screen <canvas> and its raw RGBA bytes are read with getImageData. Two complementary tests run over the R, G and B channels:

Chi-squared (Pairs of Values) test. For every pair of adjacent values that differ only in the last bit — (0,1), (2,3), … (254,255) — the test compares the observed counts against the average of the pair. In a clean image these pairs are usually unequal; sequential LSB embedding pushes each pair toward equality. A large chi-squared statistic signals likely hidden data.
LSB-balance check. It counts how many least-significant bits are 1 versus 0. Natural images are rarely exactly 50/50; a ratio extremely close to 0.5 across millions of pixels is a hallmark of an embedded bitstream, because a random message bit-stream is naturally balanced.

The two signals are combined into a plain-language verdict.

Why LSB steganography works — and why it is detectable

The human visual system cannot distinguish a pixel value of 200 from a pixel value of 201. The colour difference is smaller than the noise introduced by JPEG compression, lighting variation, or camera sensor noise. This is why encoding one bit of a hidden message in the last bit of each channel produces an image that is visually indistinguishable from the original.

However, natural photographic images are not random. Adjacent pixel values tend to cluster — they follow the texture and colour of real surfaces. The statistical distribution of value pairs like (200, 201) in a real photo is usually quite unequal. Sequential LSB embedding overwrites each channel’s lowest bit with a bit from the hidden message, which is approximately random. That randomness flattens the pair counts toward equality — and that is what the chi-squared test detects.

Limitations and false positives

Scenario	Effect on detection
Re-saved as JPEG or PNG with re-encoding	LSBs are destroyed; detection fails
Dithered or synthetic image	Naturally random LSBs may trigger false positive
Payload in only a small region	Whole-image statistic may score low despite real embedding
Non-sequential or encrypted LSB scheme	May distribute bits in patterns that evade this test
Palette-mode PNG	Colour-indexed images need different analysis

Tips for forensic use

Use the original file as downloaded or received — any conversion, screenshot, or re-save can wipe or corrupt the LSBs.
A high score warrants deeper investigation with dedicated forensic tools that visualise individual bit planes and attempt extraction with known schemes (for example, OpenStego, StegSolve, or Aletheia).
A “clean” verdict means sequential LSB embedding is unlikely to be present across the whole image — it does not rule out non-standard or partial embedding techniques.

Your image is processed entirely in your browser via the Canvas API. Nothing is uploaded.