Emirati researcher develops groundbreaking algorithm to share data without breaching privacy
An Emirati researcher has developed a groundbreaking algorithm that allows vast amounts of data to be analysed without its security being breached.
PhD student Abdelrahman AlMahmoud, 29, came up with a method of looking at specific sections of encrypted data without compromising the information as a whole.
The process – known as privacy preserving analysis – means companies can now share agreed data with one another without running the risk of it being stolen or decrypted.
The breakthrough is significant because firms will be able to analyse and learn from much bigger data sets - potentially extracting more valuable information.
“Companies wishing to analyse the data they and others hold first agree on the parameters of what they hope to achieve by studying it,” said Mr AlMahmoud.
“Once that’s done we prepare the data in a particular way that preserves the necessary properties within it to run an analysis while it stays encrypted.
“The novelty is that we don’t decrypt the data we’re sent at any point. It always remains encrypted and so always remains secure.
“There are other algorithms that do this but ours is faster and it’s using a new approach to this problem or challenge.”
Mr AlMahmoud, a student at Khalifa University and researcher at the Emirates ICT Innovation Centre, first began working on the algorithm six months ago.
Since then he and three other academics have been able to test the application in several practical scenarios, proving its effectiveness each time.
Next month, he intends to apply for a patent to protect the new algorithm, before starting the process of selling the idea internationally.
“First and foremost the system is practical,” Mr AlMahmoud said. “We’ve been able to implement it and get some very good results in real-life scenarios.
“Most other algorithms are not usable in a live application but our approach is more practical and we’re aiming to push it to the market really soon.”
Mr AlMahmoud gave an example of suspected bank fraud as an area where his algorithm can assist big business.
Currently, one bank may be reluctant to share its data with another owing to obvious privacy implications and the sensitivity of the information held.
Using his system, however, banks can agree to share very specific properties within a data set without exposing the material in full.
This can allow, for example, firms to identify fraudulent activity across the sector while keeping customers’ records encrypted and secure.
“We require a lot of data to run complex machine learning algorithms,” said Mr AlMahmoud.