Secure Digital Content Moderation Using a Hybrid AI-Based Morphed Image Detection System with Instant Removal and Quarantine

Abstract

The increasing use of virtual media has raised the authenticity of pictures shared online. This paper introduces a server-side technique for detecting morphed images before uploading the post on social media. The proposed model employs a pre-trained EfficientNetB3 version to analyse the images by extracting key capabilities such as unique Pixel value, Noise Pixel value, edges to  determine whether the image is authentic or  morphed based on confidence scores if its high-confidence then the morphed images are immediately blocked from being uploaded and the user gets  a  notification of warning for causing an illegal attempt, while borderline cases detected then the images are  quarantined for manual review to reduce false positives. Only authentic images are approved for uploading after ensuring content legitimacy. This model also tests the machine's accuracy and performance in showing how traditional photograph forensic methods is used in identifying minor alterations and outcome indicates the strength of incorporating device learning into content material moderation. It also reduces the distribution of manipulated or morphed images online. This method complements the credibility of digital content and prevents illegal activities sharing on social media