Authorship Attribution of Android Apps

摘要: Since the first computer virus hit Advanced Research Projects Agency Network (ARPANET) in early 1970s, security community interest revolved around ways to expose identities of malware writers. Knowledge adversarial promised additional leverage experts their ongoing battle against those perpetrators. At dawn computing era, when writers and malicious software were characterized by lack experience relative simplicity, task uncovering was more or less straightforward. Manual analysis source code often revealed personal, identifiable information embedded authors themselves. But these times have long gone. Modern day's extensively use numerous generators mass produce new variants employ advanced obfuscation techniques hide identities. As a result work trying uncover became significantly challenging time consuming. To gain insight into identity an adversary, we turn our attention authorship attribution research, which offers broad spectrum for identifying author document, based on author's writing style. Equipped with methods, explore Android binaries role features related development process determination binary authorship. Within this context, propose incremental approach perform apps. First set known then generation profiles unknown We assess effectiveness several sets legitimate produced actual developers, as opposed using artificially created authors' data. achieve 97.5% accuracy authors» further evaluate than 131,000 apps collected from various sources including 10 different markets globe.