Study: Deidentifying wearable data may not be enough to protect privacy

Deidentifying information from wearable units might not be sufficient to guard customers’ privateness, in line with a evaluation of research published in the Lancet Digital Health. 

The evaluation targeted on research that evaluated whether or not people may very well be reidentified based mostly on biometric indicators from wearables. The researchers included 72 research of their closing evaluation. Most targeted on utilizing EEG, ECG and inertial measurement unit (IMU) information, like utilizing a tool’s accelerometer or gyroscope to measure various kinds of motion and gait.

General, 17 research demonstrated a capability to establish a person based mostly on EEG. 5 of these research included the recording size wanted to establish customers: 21 seconds on common, with a median of 12.8 seconds. Eight research discovered a strategy to reidentify customers based mostly on ECG, whereas 13 may pinpoint people based mostly on their strolling gait. 

“In conclusion, an actual danger of reidentification exists when wearable gadget sensor information is shared. Though this danger might be minimised, it can’t be absolutely mitigated. Our findings reveal that the fundamental practices of withholding identifiers from public repositories won’t be adequate to make sure privateness,” the researchers wrote. “Extra analysis is required to information the creation of insurance policies and procedures which can be adequate to guard privateness, given the prevalence of wearable-device information assortment and sharing.”

WHY IT MATTERS

The examine’s authors discovered most of the research they reviewed had excessive right identification charges, and customers may very well be recognized with comparatively small quantities of sensor information. Nevertheless, they did notice that most of the research included within the evaluation had small teams of individuals, which may restrict its generalizability with bigger teams. Nonetheless, the 4 research with bigger populations did have comparable outcomes because the smaller research.

As extra well being information turns into extra obtainable and organizations just like the FDA and the NIH encourage its use, the examine’s authors argue researchers and information scientists might want to contemplate new methods to guard consumer privateness.

“The findings right here shouldn’t be used to justify blocking the sharing of biometric information from wearable units. Quite the opposite, this systematic evaluation exposes the necessity for extra cautious consideration of how information must be shared because the danger of not sharing information (eg, algorithmic bias and failure to develop new algorithmic instruments that might save lives) could be even better than the chance of reidentification,” they wrote. “Our findings counsel that privacy-preserving strategies can be wanted for open science to flourish. For instance, there is a chance for regulatory our bodies and funding businesses to broaden help for privacy-conscious data-sharing platforms that mitigate reidentification danger.”