/jalangi2

Dynamic analysis framework for JavaScript

Primary LanguageJavaScriptApache License 2.0Apache-2.0

Jalangi2

Introduction

Jalangi2 is a framework for writing dynamic analyses for JavaScript. Jalangi1 is still available at https://github.com/SRA-SiliconValley/jalangi, but we no longer plan to develop it. Jalangi2 does not support the record/replay feature of Jalangi1. In the Jalangi2 distribution you will find several analyses:

See our tutorial slides for a detailed overview of Jalangi and some client analyses.

Requirements

We have tested Jalangi on Mac OS X with Chromium browser. Jalangi should work on Mac OS 10.7, Ubuntu 11.0 and higher and Windows 7 or higher. Jalangi will NOT work with IE.

On Windows you need the following extra dependencies:

  • Install Microsoft Visual Studio 2010 (Free express version is fine).
  • If on 64bit also install Windows 7 64-bit SDK.

If you have a fresh installation of Ubuntu, you can install all the requirements by invoking the following commands from a terminal (package names may be out of date).

sudo apt-get update
sudo apt-get install python-software-properties python g++ make
sudo add-apt-repository ppa:chris-lea/node.js
sudo apt-get update
sudo apt-get install nodejs
sudo apt-get update
sudo apt-get install chromium-browser

Installation

Clone the repository, and then run:

npm install

Run tests

python scripts/test.traceall.py
python scripts/test.analysis.py
python scripts/test.dlint.py

Usage

Analysis in node.js with on-the-fly instrumentation

An analysis can be performed on a JavaScript file in node.js by issuing the following commands:

node src/js/commands/jalangi.js --inlineIID --inlineSource --analysis src/js/sample_analyses/ChainedAnalyses.js --analysis src/js/sample_analyses/dlint/Utils.js --analysis src/js/sample_analyses/dlint/CheckNaN.js --analysis src/js/sample_analyses/dlint/FunCalledWithMoreArguments.js --analysis src/js/sample_analyses/dlint/CompareFunctionWithPrimitives.js --analysis src/js/sample_analyses/dlint/ShadowProtoProperty.js --analysis src/js/sample_analyses/dlint/ConcatUndefinedToString.js --analysis src/js/sample_analyses/dlint/UndefinedOffset.js tests/octane/deltablue.js

In the above analysis, we chained several analyses by including --analysis src/js/analyses/ChainedAnalyses.js as the first analysis. The command runs the following analyses

src/js/sample_analyses/dlint/CheckNaN.js
src/js/sample_analyses/dlint/FunCalledWithMoreArguments.js
src/js/sample_analyses/dlint/CompareFunctionWithPrimitives.js
src/js/sample_analyses/dlint/ShadowProtoProperty.js
src/js/sample_analyses/dlint/ConcatUndefinedToString.js
src/js/sample_analyses/dlint/UndefinedOffset.js

The implementation of an analysis requires the implementation of several callback functions. One can start writing an writing analysis using the template file src/js/runtime/analysisCallbackTemplate.js. A documentation of these call back functions can be found at docs/MyAnalysis.html. A tutorial on writing a Jalangi analysis can be found at docs/tutorial1.md. While writing an analysis one could run src/js/sample_analyses/pldi16/TraceAll.js analysis on a JavaScript file to print all the callback functions that got called during the execution of the file. Such a trace is useful to see what callbacks get called during an execution. The following command runs the TraceAll.js analysis on the file tests/octane/deltablue.js.

node src/js/commands/jalangi.js --inlineIID --inlineSource --analysis src/js/sample_analyses/ChainedAnalyses.js --analysis src/js/runtime/SMemory.js --analysis src/js/sample_analyses/pldi16/TraceAll.js tests/octane/deltablue.js

Analysis in node.js with explicit one-file-at-a-time offline instrumentation

An analysis can be performed on a JavaScript file in node.js by issuing the following commands:

node src/js/commands/esnstrument_cli.js --inlineIID --inlineSource tests/octane/deltablue.js
node src/js/commands/direct.js --analysis src/js/sample_analyses/ChainedAnalyses.js --analysis src/js/sample_analyses/dlint/Utils.js --analysis src/js/sample_analyses/dlint/CheckNaN.js --analysis src/js/sample_analyses/dlint/FunCalledWithMoreArguments.js --analysis src/js/sample_analyses/dlint/CompareFunctionWithPrimitives.js --analysis src/js/sample_analyses/dlint/ShadowProtoProperty.js --analysis src/js/sample_analyses/dlint/ConcatUndefinedToString.js --analysis src/js/sample_analyses/dlint/UndefinedOffset.js tests/octane/deltablue_jalangi_.js

In the above analysis, we chained several analyses by including --analysis src/js/analyses/ChainedAnalyses.js.

Analysis in a browser using offline instrumentation

An analysis can be performed on a web app using the Chrome browser by issuing the following commands:

node src/js/commands/instrument.js --inlineIID --inlineSource -i --inlineJalangi --analysis src/js/sample_analyses/ChainedAnalyses.js --analysis src/js/sample_analyses/dlint/Utils.js --analysis src/js/sample_analyses/dlint/CheckNaN.js --analysis src/js/sample_analyses/dlint/FunCalledWithMoreArguments.js --analysis src/js/sample_analyses/dlint/CompareFunctionWithPrimitives.js --analysis src/js/sample_analyses/dlint/ShadowProtoProperty.js --analysis src/js/sample_analyses/dlint/ConcatUndefinedToString.js --analysis src/js/sample_analyses/dlint/UndefinedOffset.js --outputDir /tmp tests/tizen/annex
open file:///tmp/annex/index.html

While performing analysis in a browser, one needs to press Alt-Shift-T to end the analysis and to print the analysis results in the console.

Analysis in a browser using a proxy and on-the-fly instrumentation

You can also setup a proxy to instrument JavaScript files on-the-fly. To do so, you need to install mitmproxy. We have tested mitmproxy version 7.0. On Linux, you can follow the standard installation instructions, but instead of running sudo pip install mitmproxy, run sudo pip install mitmproxy==7.0.0 to get the tested version. On Mac OS, the easiest path we have found is to use Homebrew. With Homebrew installed, you can install the right version by running:

brew install python
pip install -U pip
pip install mitmproxy==7.0.0

Note that you might need to restart your shell afterward, to ensure the python being used is /usr/local/bin/python.

For instrumenting code served over HTTPS, you will additionally need to set up a root certificate for mitmproxy. See their instructions or this document.

After installation, you can run the Jalangi instrumentation proxy by issuing the following command:

mitmdump --quiet --anticache -s "scripts/proxy.py --inlineIID --inlineSource --analysis src/js/sample_analyses/ChainedAnalyses.js --analysis src/js/runtime/analysisCallbackTemplate.js"

In your browser, the http and https proxy should be set to 127.0.0.1:8080. Now if you load a website in your browser, all JavaScript files associated with the website will get instrumented on-the-fly.

On a Mac, the proxy can be set and launched automatically by issuing the following command:

./scripts/mitmproxywrapper.py --toggle --auto-disable --quiet --anticache -s "scripts/proxy.py --inlineIID --inlineSource --analysis src/js/sample_analyses/ChainedAnalyses.js --analysis src/js/runtime/analysisCallbackTemplate.js"

The command starts mitmproxy if the proxy is not currently enabled, and disables it otherwise. The --auto-disable option will automatically disable the proxy when the script is interrupted.

Jalangi2 caches the instrumented source files in ./cache/. The use of the cache can be disabled during development by passing the --no-cache flag to scripts/proxy.py.

Developing an analysis in Jalangi2

Refer to docs/index.html and docs/commands.md for further information. A tutorial on writing a Jalangi analysis can be found in docs/tutorial1.md.

Supported ECMAScript versions

Jalangi2 supports ECMAScript 5.1. Some ES6 features may work, but have not been tested.

License

Jalangi2 is distributed under the Apache License.