Image spam — ASCII to the rescue!

We take an unorthodox approach to image spam detection, by applying existing software and decades-old technology: ASCII art. Our technique is straightforward and gets good levels of detection over a corpus with 1159 ham and 1492 spam images, with a tolerable amount of misclassifications. Furthermore, we only look at the images themselves, meaning that this method can be trivially enhanced by combining it with existing anti-spam techniques.