The researchers analyzed samples from 120 people enrolled in the Human Microbiome Project, a government-sponsored consortium to study the human microbiome. Using samples of microbes from their skin, their mouths, and other parts of the body, along with stool samples, the researchers created a metagenomic profile for each volunteer.

Anywhere from a month to a year later, the volunteers came back donated more samples. The researchers created a second metagenomic profile for each individual, and then they tested whether they could match the profiles across time.

Using microbiome samples from the skin, mouth, and vagina, the researchers’ algorithm correctly identified about 30 percent of participants in the study. That’s not a good enough success rate for forensics, but it is enough to raise concerns about the privacy of participants in studies on the human microbiome. Those samples are connected with other personal data, such as information about age, area of residence, illnesses, and general health status. If there’s a one in three chance that a person’s microbes can identify them, microbiome databases may be more sensitive than anyone thought.