AI breakthrough helps detect hidden apps on smartphones

Study finds machine learning can detect hidden vault apps on Android with up to 98% accuracy, aiding digital forensics

Listen to article

Researchers in Australia have developed a powerful new technique to uncover hidden "vault" apps on smartphones — a discovery that could aid law enforcement in digital investigations.

The study, conducted by Edith Cowan University and the University of Southern Queensland, found that machine learning (ML) can identify vault apps with up to 98% accuracy on Android devices.

Vault apps allow users to store files, messages, or even other apps behind layers of encryption. While often used for privacy, they have increasingly been linked to illicit activities, including espionage and unauthorised surveillance.

“These apps can mimic normal ones, making them very difficult to detect,” said Associate Professor Mike Johnstone from ECU. “Current detection tools rely on prior knowledge of suspicious apps, which limits their usefulness.”

By contrast, the new machine learning approach can identify vault apps without needing a pre-existing list or database.

The breakthrough could offer a valuable tool for police and security agencies, particularly as smartphones become more integral to modern life, with over 5 billion users worldwide.

“Given how common smartphones are, any non-invasive and accurate method for identifying these hidden apps could be a game-changer,” said Professor Johnstone.

The team now plans to expand the research to include more algorithms, a wider dataset, and tests on non-Android devices.