Twitter Data Analysis to Enhance Malware Detection Using ML

Abstract

In current times, we have seen an escalation of cellular or apps, consisting of useful, congenial apps, and additionally malicious apps (or malware). Detecting fraudulent applications is a challenging but critical task, as malicious apps may cause significant damage and financial losses to their users. The majority of strategies for detecting malware rely on characteristics acquired from the apps’ code via static or dynamic analysis. Much zero-day malware software, on the other hand, avoids such mechanisms and enters the market. We recommend using social media statistics, specifically, Twitter, to supplement the statistics contained within the code and facilitate the detection of zero-day android malware apps. We recommend picking out tweets that mention android malware, primarily those that can contribute to the malware’s spread. The notion is that clients who try to sell it and/or spread malware share the same characteristics as spammers. We utilized the Twitter Developer APIs to scan a huge number of tweets that had URLs that were similar to those found in android apps. The tweets were recorded in a MongoDB database, together with meta-statistics about their retweets/favorites and customers. The URLs found throughout the stream of tweets were matched with android apps using data gathered from the Google Play Store. Furthermore, utilizing a platform known as AndroZoo, which uses antivirus tools such as VirusTotal to detect malware, the apps identified in tweets that were linked to apps in the Google Play Store were classified as benign or malicious. Furthermore, Twitter users who post malware are being investigated to uncover tendencies similar to those seen in unsolicited mail, which might be used to identify zero-day malware.