The Google Play Store has millions of applications already with thousands being uploaded every month. This makes it somewhat difficult for Google to filter between the good apps and the bad. However, the company is now using machine learning in the form of peer grouping to filter malicious apps.

Google achieves this by comparing a new application with a bunch of pre-approved apps to check if there are any new permissions required. So if an app is asking for unusual permissions that other identical apps don’t, the new system flags it immediately and the company’s engineers have a closer look at them.

Since broadly grouping the apps based on their categories isn’t feasible, the algorithm groups the apps based on their description, file sizes, and some other data. Some of these measures are doing wonders for the Android platform already according to Google’s recent security review. It showed that the number of users who downloaded malicious apps on Android were down from 0.15% percent in 2015 to 0.05% percent last year.

But the number is significantly higher when we take apps downloaded from other app hubs into account. In most regions of Asia, particularly in China, most users download the majority of their apps from third-party app stores. In this case, the percentage of users who downloaded malicious apps rose from 0.5% in 2015 to 0.7% in 2016. It’s clear that Google will have to come up with an alternative strategy to combat the menace of malicious apps that are making their way to the devices via alternative app hubs.

“We focus on signals that can negatively affect user privacy, such as permission requests that are not related to core app functionality, and the actual, observed behaviors,” said Google’s Martin Pelikan in an email to The Verge. “For example, a flashlight app might not need access to address book of the user or the precise hardware identifier of a user’s phone. The same might hold for many other apps, such as ‘mirror’ apps that turns on a device’s front-facing camera.”