Researchers have recently discovered a novel way some threat actors are tricking these scanning engines, and this bot avoidance mechanism in particular has been deployed on multiple Office 365 credential phishing websites. Image recognition software is improving and becoming more accurate, this new technique aims to deceive scanning engines by inverting the colors of the image, causing the image hash to differ from the original. This technique can hinder the software’s ability to flag this image altogether.
However, a victim visiting the website would likely recognize that the inverted picture is illegitimate and exit the website. As a result, the threat actor has stored the inverted image and, within the index.php code, has used a CSS method to revert the color of the image to its original state.