Taking the text out of the Spam

For spammers, the trouble with image and video spam is that they have to ultimately give you information. Selling Viagra – you need to know where to buy it. Stock pump and dump – you need to know what stock you need to buy. So, leaving audio based spam aside for now, the text has to be given to you and it has to be readable, or it’s of no use to the spammers. Text-based spam content-filtering has come a long way, so as long as we can extract the text from the images and video, we should be able to run that text through the existing text filters.

Many top websites use “Captchas”. This is distorted text, designed to be unreadable by computers (used in automation by spammers), but is readable by humans. This is exactly what spammers are trying to do with image-spam. Whilst websites are using distorted image text to stop spammers, the spammers are using distorted image text to bypass email spam filters. The irony is that as spammers seem to be using social engineering and a little ingenuity to defeat catchas, image-based spam filters are still struggling. So what if we used spam images as captchas? Could we somehow use this to get spammers to unwittingly convert their images back to text?