Clickability tracking pixel

Missouri Researchers Use Machine Learning to Secure IoT

Researchers with the Missouri University of Science and Technology are hoping to use federated learning to train algorithms in an attempt to secure devices that make up the Internet of Things.

by The Rolla Daily News / November 18, 2020
Shutterstock

(TNS) — The answer is machine learning, according to researchers at Missouri University of Science and Technology.

Missouri S&T researchers want to ensure that Internet of Things (IoT)-collected data is accurate and usable, while still protecting the items from malicious attacks or invasions of privacy. IoT are physical objects with sensors and software that are connected to the Internet. Researchers say that improving a machine-learning technique called federated learning could allow companies to develop new ways to collect anonymous, but accurate, data from users.

Federated learning trains algorithms with access to multiple individual devices that hold local data. Federated learning doesn't exchange data with the items, which means there is no central dataset or server where it all the information is stored. With the lack of shared data in federated learning, concerns such as privacy, security and access rights could become a non-issue.

"Federated learning is a game changer for IoT because it enables machine learning without needing the learner to directly access customer data," says Dr. Sajal Das, the Daniel C. St. Clair Chair of computer science at S&T. "IoT provides a fertile ground for applying federated learning to private devices that are rich in data."

Das warns that IoT devices are vulnerable to dynamic environments and attacks from outside sources with erroneous data. Therefore, he says collecting data in a federated manner is crucial.

Das and his co-investigator Dr. Tony Luo, an associate professor of computer science at S&T, are designing new federated learning algorithms with funding from the National Science Foundation and are putting data safety and accuracy above all else in their work.

"By collecting data from numerous IoT devices without compromising privacy or network capabilities, our methods will allow for growth in the way these devices work and measure data," says Das. "Our new algorithms will combat erroneous data by designing novel incentive mechanisms to motivate and encourage users who contribute accurate data."

Das and Luo hope that users will be willing to contribute data to machine learning while having confidence that the data is not identifiable. That way, the data can be used to push the boundaries of complexity and performance for IoT items.

Das says that the research has the potential to produce tremendous benefits to personalized industries such as health care.

"In smart health care, wearable IoT devices can help measure an individual's health conditions such as vital records, physical activities and food intake," says Das. "For example, without directly accessing a patient's sensitive and private information, our novel federated learning approach can investigate how diseases like diabetes are influenced by lifestyle and demography and whether there is correlation with other health conditions like hypertension."

Das says that with enough advances in the secure and accurate collection of IoT data, new devices could serve more and better purposes while easing the minds of those who are reluctant to accept smart technology into their homes.

©2020 The Rolla Daily News, Distributed by Tribune Content Agency, LLC.

Looking for the latest gov tech news as it happens? Subscribe to GT newsletters.

E.REPUBLIC Platforms & Programs