Artificial intelligence (AI) usually takes advantage of machine learning, a field of computer science that enables computers to think and learn similarly to humans – based on examples, without programming. So what if we combined the power of AI with data safety? A team of researchers at the University of Helsinki, Finland, has recently done just that. They developed a new privacy-aware machine learning method that keeps data private and secure.
Most sensitive data needs to be private and safe, but this is especially true with data used in applications concerning human behavior and health. However, current security and privacy methods don’t allow for complete privacy – one party with unrestricted access to all data is always needed.
Things may change with the new privacy-aware machine learning methods. These approaches are based on the concept of differential privacy, meaning they can reveal only limited information on each data subject. Here is how Assistant Professor Antti Honkela of the University of Helsinki explains it: “Our new method enables learning accurate models, for example, using data on user devices without the need to reveal private information to an outsider.”
Interestingly, the new method is not only applicable to data protection – the team also used it to predict cancer drug efficacy using gene expression. And unlike other AI-based privacy models, this method is able to learn from smaller data, which makes everything easier.
Reference:
University of Helsinki (https://www.helsinki.fi/en/news/data-science/new-ai-method-keeps-data-private)
Chanwoo's Weekly Tech 