Image spam is gaining popularity with spammers because most spam filters are not sophisticated enough to identify spam that's been embedded in a photograph or image. To combat this, several researchers around the country have been working on highly accurate methods to detect image spam. Ultimately, the research will be incorporated into individual or corporate email systems to further reduce the incidence of spam.

Researchers at the University of Pittsburgh have developed a technique that can identify spam messages embedded in an image. The technique predictively evaluates simple image properties to identify spam. The success rate of the Pittsburgh algorithm is 90-99% for inbound spam.

Princeton University scientists are also targeting hidden spam. According to the researchers, image spam is often sent in batches. The files differ only because the spammers have integrated randomization information in the spam. The Princeton filter identifies suspected image spam, then compares it to other similar messages. If the algorithm detects a substantially similar file, the messages are flagged as image spam. This technique has a less than 0.001 false-positive rate and a high spam detection rate.

Georgia Tech has also developed a similar classification system that compares four different properties of a message to determine the likelihood that the message is spam. Suspected image spam is scored and discarded if the spam threshold is exceeded. This method is accurate more than 80 percent of the time.