LibFP - Library Fingerprinting Framework

Recent years have witnessed rapid development in the detection of third-party libraries in Android applications. However, each approach typically comes with its own implementation, limiting re-use for further development. To address this LibFP was created - a versatile third-party Library Fingerprinting Framework written in Java.

Installation and Build

This project uses Gradle as its build system. Follow these steps to build and install the LibFP framework:

Clone the repository:

git clone https://github.com/example/libfp.git

Navigate to the project directory:
```
cd libfp
```
Build the project using Gradle:
```
gradle build
```
After a successful build, you can find each compiled JAR file in their corresponding build/libs directory. Note that in order to utilize libfp-cli as an executable, run
```
gradle libfp-cli:standalone
```

Requirements

Java 17+
Android framework JAR (android.jar) - stubs file is ok
Dataset from LibLOOM formatted according to the Demo/ directory (see dataset.conf.

Modules

Module	Description
libfp-core	Core module providing fundamental components for profile generation and third-party detection.
libfp-cha	Module implementing the Class Hierarchy Analysis (CHA) strategy for creating library profiles.
libfp-benchmark	Module containing benchmarking functionalities, including test suites and result analysis.
libfp-tlsh	Module integrating the Trend Micro Locality-Sensitive Hashing (TLSH) algorithm for similarity.
libfp-impl	Implementation module for additional features, extensions, and custom profile generation.

Example Code Snippet

Here's a simple example demonstrating how to use the LibFP framework to perform benchmarking:

// Import necessary classes and interfaces
// ...
// 1. retrieve the dataset
DataSet dataset= //...;

// 2. Create the TestSuite with all relevant factories. Please follow
// the steps provided in CustomProfileExample to create
// ProfileManager and Strategy instances.
ICHAProfileProvider context = new ExampleProfileContext();
CHATestSuite suite = new TestSuiteBuilder<>(CHATestSuite::new)
        .setDataSet(dataSet)
        .setProvider(context)
        .createTestSuite();

// 3. Specify the app name(s) to test.
String appName = "003-nodomain.freeyourgadget.gadgetbridge";

// 3.5. (optional) Prepare the profiles.
suite.prepareApp(appName);
suite.prepareLibraries();

// 4. Benchmark a strategy by using a custom threshold configuration
IThresholdConfig config = new SimpleThresholdConfig()
        .set(CHAProfile.class, 0.8)   // profile similarity threshold
        .set(ClassProfile.class, 0.6) // class similarity threshold
        .set(TestResult.class, 0.68); // result threshold
        
BenchmarkResult result = suite.benchmark(appName, config);
// 5. Retrieve the accuracy of the chosen strategy by providing the
// appType (a default type is "").
Whitelist whitelist = dataSet
        .groundTruth()
        .getVersionWhitelist(appName);

TestAccuracy accuracy = result.getTestAccuracy(
        BenchmarkResult.defaultAppType, config, whitelist);
// ...

Additional examples are placed in libfp-impl:example/.

License

Distributed under the MIT license. See LICENSE for more information.

MatrixEditor / libfp