This is the Proof-of-Concept implementation of SCAnDroid, a tool published at WiSec 2018. See the "SCAnDroid" paper by Spreitzer, Palfinger, and Mangard for more details.
This repository contains a tool to analyze the Android API for possible information leaks. SCAnDroid relies on the concept of template attacks. Thus, it works without prior knowledge about possible information leaks.
The following figure illustrates the design of SCAnDroid, which consists of an Android app (SCAnDroid) and the Backend. The Backend consists of a tool to fetch and parse packages of the Android API, a tool to control the Android app as well as the Android device via the Android Debug Bridge (ADB), and a tool to analyze the gathered information. The Analyzer and the Controller component are based on the "ProcHarvester" framework.
The basic work flow is as follows:
- Fetch packages: The Parser component fetches a list of available constructors, and methods from the Android Developers website.
- Parse: The Parser component parses the information to extract a list of methods to be profiled.
- Trigger event: The Controller component triggers events, such as website launches, Google Maps search queries, or app starts, on the device.
- Log: The Android app (SCAnDroid) relies on Java Reflections to create all objects and profiles/invokes Android APIs, while events are being triggered.
- Fetch data: After the profiling phase, the log files are fetched to the Backend.
- Analyze: The log files are analyzed to identify information leaks that allow inferring the triggered events.
We used SCAnDroid to analyze information leaks that allow inferring the following events from Android APIs:
- Website launches
- Google Maps search queries
- App starts
Warning: This code is provided as-is. You are responsible for protecting yourself, your property and data, and others from any risks caused by this code. This code may not detect vulnerabilities in your application/OS or device. This code is only for testing purposes. Use it only on test systems which contain no sensitive data.
