Researchers have developed a new, lightweight method for detecting image forgeries using only CPU computation and gradient boosted trees. This approach employs multi-scale Error Level Analysis (ELA) across various JPEG quality levels, along with novel cross-quality ELA ratio features to identify double-compression artifacts. When evaluated on a format-controlled subset of the CASIA v2.0 dataset, the method achieved an AUC of approximately 0.990 and an F1 score of 0.905, demonstrating its effectiveness in detecting compression-history inconsistencies rather than relying on file-format shortcuts. AI
IMPACT Offers a lightweight, CPU-only alternative for digital forensics, potentially enabling wider deployment of forgery detection tools.
RANK_REASON The item is an academic paper detailing a new method for image forgery detection. [lever_c_demoted from research: ic=1 ai=0.4]
AI-generated summary · Google Gemini · from 1 sources. How we write summaries →
