top of page

Android Malware Detection using Machine Learning

Updated: Mar 1, 2020


Android is a working framework essentially focusing on portable gadgets, for example, cell phones and tablets. The working framework has held a vast offer of the market for a couple a long time now, and its piece of the pie keeps on developing. Much the same as numerous other cell phone working frameworks, Android's vital moving point is countless applications or applications for short that are created by different gatherings. The Web availability on Android-based gadgets gives these gadgets and going with applications practically boundless routes for cooperation with other PC frameworks on the Internet.

These certainties additionally make Android an engaging focus for vindictive programming (malware). A client may introduce a malware application on an Android gadget without realizing that the application is vindictive. A few instances of malignant activities are sending unapproved SMS messages, and sending delicate private data put away in the gadget to a remote server without the client's learning and consent. So as to moderate dangers presented by a malware application, we propose a totally programmed system to distinguish vindictive conduct.

The structure comprises of two essential parts: testing, what's more, machine realizing. The the testing part is there to altogether test each application in a controlled situation altogether to learn however much as could be expected about its conduct. The more intensive testing of the application can be completed, the better include vector catching the genuine conduct of the application can be built. The machine learning segment serves the motivation behind separating these component vectors from Android applications and afterward utilizing the element vectors to learn parameters of a classifier that segregates amiable from vindictive applications.


For differentiating the Benign Android apps from the malicious android app, we first need to upload the APK in the system after that it will be decompiled for further feature extraction. And after feature extraction, it is used in the Random Tree classifier technique. The complete approach is shown in the following diagram:


This contains following broad blocks in this project:

Android Application: This is used as an interface to upload the test APK on the server and then show the result after running through the trained model.

Server: This is the place where actual logic runs. It decompiles the APK and run the trained model and gives output if there is any malware present in the uploaded APK.

Machine Learning Model: The model is trained with multiple APKs and contains a list of features and learned parameters that are used for checking the uploaded test APK.

Test APK: APK used for testing if the model is giving correct output. This contains two types of applications, one with malware and one without malware.

Reach out to us for more details about this project. Fill the form on our website ( or directly connect with us on call/WhatsApp +91-9545954177.


Technical Team Member

34 views0 comments


bottom of page