Basic text on a single color background is easy to read, but draw a line through it or add additional data into the background and it will crap out, like contrasting greens or lines through the text. I am one of those who uses graphics to hide email adresses from automated bots or spiders. Until now, it would take a human eye to harvest the e-mail address and start spamming. Less likelely considering the are billions of web sites. So this virus writer takes that idea and uses it to disquise data that could otherwise be easily picked up by text scanners and tries to turn the book on them. But sophos picked up on it and wrote some code to pull it out of a graphic file. I wonder if it's traditional OCR technology or something along the lines of looking at the code in terms of binary not optical. To use traditional OCR definition in this case is wrong because the software isn't looking at it from a optical standpoint. It's not OCR they are using, is my point?