Error Level Analysis for Image Forensics

JPEG stores an image using lossy compression, so saving it again can change pixel values. Error level analysis makes one such change visible by recompressing a JPEG at a chosen quality and displaying the difference from the input. A 2015 evaluation studied ELA under various tampering and compression conditions; broader JPEG-forensics research examines traces of compression and quantization. These methods can be useful in a controlled investigation, but a colorful ELA display is not a verdict about authenticity. Interpretation depends on the file’s history. A picture may have been resized, edited for exposure, uploaded to a platform, downloaded and re-saved before an analyst sees it. Text overlays, high-detail areas and different local textures can also change residual appearance. A region that looks unusual may reflect ordinary processing rather than a pasted object. Conversely, a manipulated image can be saved consistently enough that a simple ELA view offers little contrast. The chosen recompression quality affects the result. Start with the highest-quality original available and record every transformation known to have occurred. Compare suspected regions with similar textured areas, not only with flat sky or walls. If a clue persists, inspect other evidence: visual geometry, metadata when available, source-page history and reverse-image search results. These checks also have limits. An AI-generated image may have one uniform compression history, while an authentic photo with a later text label may show different errors. For public claims, avoid phrases such as 'ELA proves this is fake.' State what file was analyzed, what comparison was made and what remains uncertain. Preserve the input and settings so another analyst can reproduce the display. ELA is a narrow compression-response tool; verification of the event or scene requires a broader evidence trail.