MALWARE DETECTION A FRAME WORK FOR REVERSE ENGINEERED ANDROID APPLICATIONS THROUGH MACHINE LEARNING ALGORITHMS

Abstract

Today, Android is one of the most used 
operating systems in smartphone
technology. This is the main reason, Android 
has become the favorite target for hackers 
and attackers. Malicious codes are being 
embedded in Android applications in such a 
sophisticated manner that detecting and 
identifying an application as a malware has 
become the toughest job for security 
providers. In terms of ingenuity and 
cognition, Android malware has progressed 
to the point where they're more impervious to 
conventional detection techniques. 
Approaches based on machine learning have 
emerged as a much more effective way to 
tackle the intricacy and originality of 
developing Android threats. They function by
first identifying current patterns of malware 
activity and then using this information to 
distinguish between identified threats and 
unidentified threats with unknown behavior. 
This research paper uses Reverse Engineered 
Android applications’ features and Machine 
Learning algorithms to find vulnerabilities 
present in Smartphone applications. Our 
contribution is twofold. Firstly, we propose a 
model that incorporates more innovative 
static feature sets with the largest current 
datasets of malware samples than 
conventional methods. Secondly, we have 
used ensemble learning with machine
learning algorithms such as AdaBoost, SVM, 
etc. to improve our model's performance. Our 
experimental results and findings exhibit 
96.24% accuracy to detect extracted malwarefrom Android applications, with a 0.3 False 
Positive Rate (FPR). The proposed model 
incorporates ignored detrimental features 
such as permissions, intents, API calls, and so 
on, trained by feeding a solitary arbitrary 
feature, extracted by reverse engineering as 
an input to the machine.