Hand writing analysis, though an imperfect practice, has proven to be a useful tool in determining identity and weeding out counterfeit documents. Researchers at the Center for Information Technology Policy
have taken the idea one step further, and developed a way to determine people's identity not from their hand writing, but how they fill in bubbles
on tests or ballots.
The crux of the research is that the stroke patterns people use to fill in the bubbles on scantron-like forms carries unique information. To test their theory, the researchers used analytical software to look at some 804 features of how each bubble was filled. Will Clarkson
, one of the researchers, wrote on his blog about the process used in the experiment.
To test the limits of our analysis approach, we obtained a set of 92 surveys and extracted 20 bubbles from each of those surveys. We set aside 8 bubbles per survey to test our identification accuracy and trained our model on the remaining 12 bubbles per survey. Using image processing techniques, we identified the unique characteristics of each training bubble and trained a classifier to distinguish between the surveys’ respondents. We applied this classifier to the remaining test bubbles from a respondent.
According to Clarkson, if people were truly random in how they filled in each bubble, their trained analytical software would have only been correct 1 out of 92 times. Instead, they claim to have achieved a 51% success rate.