Photo Forensics Tool (Error Level Analysis)

Photo forensics looks for traces that editing leaves in a JPEG file. The main test is error level analysis (ELA): the photo is saved again at a known quality and the difference is amplified, because pasted or re-saved areas often recompress differently from the rest. The tool also estimates the JPEG quality, lists compression details and editing-software tags, and compares the camera’s embedded EXIF preview with the main picture. It all runs in your browser.

How to use this tool

1. Choose a JPEG photo — ideally the original file, not a screenshot or a copy from a social network.
2. Keep the resave quality at 90% and the boost at 20×, or adjust them to see more or less detail.
3. Compare similar surfaces in the ELA image: they should have a similar brightness.
4. Read the quality estimate, the metadata and the preview check, keeping in mind that these are hints.

Formula

ELA pixel = |original − resaved at q%| × boost (per colour channel); JPEG quality = the q whose standard table matches: entry = ⌊(base × S + 50) ÷ 100⌋ with S = 5000 ÷ q below 50, 200 − 2q from 50

The file’s luminance table is compared with the libjpeg (IJG) reference table scaled for every quality from 1 to 100; cameras with their own tables get the nearest value. The preview check shrinks the main picture to the preview’s size, ignoring black bars that cameras add when the preview has a different shape and trying every rotation and mirror (browsers differ in applying the EXIF orientation). Both are smoothed with a 3 × 3 average so different resizing methods do not count, and the check flags a mean difference above 12 levels, or more than 1% of pixels differing by over 40 levels per channel.

Example

A JPEG whose first luminance entry is 3 matches the standard table at quality 90: S = 200 − 2 × 90 = 20 and ⌊(16 × 20 + 50) ÷ 100⌋ = ⌊3.7⌋ = 3. Resaving it at 90% changes most pixels by only a level or two, so the ELA image is mostly dark; a pasted area saved earlier at quality 70 would change more and look brighter.