Machine learning for mobile malware analysis

Abstract

Mobile devices are becoming more and more attached part to the majority of human’s lives, substituting computers for the use of the Internet by permitting operators to run through emails, reach banking services online, use social media at fastest ever speed through WhatsApp, Facebook, Instagram, LinkedIn, Twitter, etc. Likewise, the fast emergent attractive and supportive applications in mobile devices with irresistible experience, for example, GPS mapping and various other delivery and taxi apps using GPS for locational updates, payment transfer, and personal valet generation like Dominos, Ola, Uber, Swiggy, and Zomato make mobiles extra likeable and engaging to users. During consistent and repetitive utilization of mobiles, confidential and sensitive data like banking passwords, debit card, credit cards, contact details, and further personal data remains stored on most of the mobile devices. Based on this new set up, hackers have diverted there attention towards mobile devices as ample of wish data is available there. Existing safety software offers limited solutions against these threats and hence proving incapable in maintaining speed and delivering results with respect to express advancement in the malware industry. Hackers plan implantation of various malicious software variants like virus or spyware. Malware is a specific code developed by cyber attackers and it acts as a shorthand of malicious software. It is aimed to create broad mutilation to system, software, and data to achieve unsanctioned admittance in the network. The easiest and prevalent means of delivering malware in a mobile set is in the form of a file, link, email, or unauthorized websites. ML has already started advancement in malware detection by utilizing several types of networks, data on host, and several other anti-malware components. During the detection of malware, a formerly unnoticed sample can be a fresh file. Its secreted stuff can be malware (malign) or benign (legitimate). ML follows a wide range of methods to identify malware instead of a solo technique. These methods have various abilities and diverse responsibilities which they suit superlatively. Hence, ML may be termed as an exemplar that refers to learning from experience (that in our case is former mobile data) to advance forthcoming enactment. The solitary emphasis of this field is spontaneous learning techniques. Learning means alteration or upgrading of algorithm automatically based v on previous “experiences” deprived of any outside support from the human. As on date ML seems to be the best tool which is sharpening itself on its own to counter serious and new threats of malware in mobile devices. Hence based on rapid learning ML helps in avoiding similar nature malware attacks and also reacts to varying behavior. In our work we have undertaken compilation of various types of malwares prevalent in mobile industry, malware recognition and detection methods. An organized and inclusive gestalt of progression of malware detection techniques based on ML and the latest study on ML for mobile malware analysis is presented.

Topics & Concepts