Image privacy analyzer — check image metadata before you share it

The risk

What a single image can leak

Metadata is the part most people know about. EXIF records the camera or phone model, the exact second the shutter fired, and often GPS coordinates precise to a few meters — enough to place a home, a workplace or a child's school. XMP and IPTC fields add author names, copyright lines, document IDs and a full editing history. AI-generated files carry a second layer: prompts, negative prompts, seeds and model names, and sometimes signed C2PA content credentials that describe every production step.

The pixels themselves leak just as much. Faces of bystanders, a license plate, a whiteboard in the background, an ID badge, a reflection in a window — none of that shows up in a metadata scan. Screenshots can carry notification previews, open tabs and usernames.

It also matters where the file goes. Large social platforms strip some EXIF on upload, but images sent by email, messaging apps, cloud links or marketplaces usually arrive byte-for-byte intact. Checking image metadata before sharing takes seconds, and it is the only way to know what you are actually publishing.

Under the hood

How error level and bit-plane analysis work

JPEG compression divides an image into 8×8 blocks and rounds off detail according to a quality setting. When a file is re-saved, every block is quantized again. Error Level Analysis exploits this: re-encode the image at a known quality and measure how far each pixel moves. A picture with a uniform history shows even, low-level error. Regions that were pasted in, retouched or saved at a different quality respond differently and stand out after the 15× amplification used above.

The LSB view looks at something else entirely. Each color channel stores values from 0 to 255; flipping the lowest bit changes a value by at most one step — invisible to the eye. Steganography tools hide payloads in exactly that bit. Extracting the least significant bit of the red, green and blue channels as black-and-white planes turns that hidden layer into something you can inspect: pure noise is ordinary, while blocks, stripes or readable shapes suggest embedded data.

Both techniques inspect the pixels rather than the metadata — the same layer that gives every file the image fingerprint, its identity at the pixel level.

Honest caveats

The limits of browser-based forensics

None of these checks is a lie detector, and the results deserve careful reading. A bright patch in ELA can come from a legitimate re-save, a format conversion or a social network's recompression — not just from editing. Structure in an LSB plane can appear from sensor noise and heavy JPEG artifacts without any hidden payload. OCR depends on text size and contrast, and a compact face model misses small, turned or occluded faces.

Absence of evidence cuts the same way: a clean report does not prove an image is safe to share, and a flagged region does not prove manipulation. Treat this analyzer as a fast, private first pass. And when the metadata report shows fields you do not want to publish, the MetadataRemover metadata cleaner removes them in one pass — also entirely in your browser.