A Bayesian approach to privacy enforcement in android system
Lee, Kang En
Date of Issue2016
School of Electrical and Electronic Engineering
In recent years, the Interprocess IPC Communications and high level semantics of Android Architecture have been rendered vulnerable to privacy breaches. The existing Android Security Architecture and whitelisting features are no longer resilient to malicious native Java code execution and the manifestation of superuser privileges in invocation of dangerous Android system permissions calls. This thesis studies in depth the use of dynamic taint analysis for behavioral reconstruction of common Android Trojan and Malwares. This problem is then dissected using Bayesian data mining, leveraging on the Bayesian classifier which uses supervised learning and statistical correlation to impact a good performance latency. Two custom Java implementation frameworks are proposed and designed in this project. One is a lightweight Bayesian Android application that is agnostic to the Android runtime system, to analyze network packet data. The other implementation is an insightful granular approach using a Java platform, to filter Android SMS sent from the Android Binder Service into Spam and Ham categories. The frameworks are tested for robust runtime accuracy and reduced overhead.
Final Year Project (FYP)
Nanyang Technological University