Neural networks and automated learning were used to define an algorithm to search for similar text descriptions and visually similar icons for the most common applications in the application store. The algorithm detected 49,608 potential threats.
According to a two-year study, 2,040 applications were classified as harmful, while some required a suspicious number of permissions to access user information.
Of the 1.2 million applications accessed, a large number of counterfeit applications requested access to user data, including Hill Climb Racing and Temple Run.
Of these applications, about 7,046 were malicious, including 2,040 “counterfeit and high risk”.
A separate report from Forbes magazine indicated that threats posed by mobile phone applications still threatened Android phone users and the iOS system.
Google and Apple are working to strengthen their systems against cybercriminals.
Google also noted that the number of rejected apps had increased by more than 55 percent over last year, while the number of apps dropped had increased to 66 percent.