README.md

Bluetooth worm simulator

This application accompanies the paper "Modelling the propagation of modern Bluetooth worms" and implements an agent-based simulator for the propagation of worm malware that spreads among personal smartphones via Bluetooth.

The simulator allows experimentation with different mobility, identification and infection models that can be independently configured or swapped out.

Installation

The simulator can be build using GHC and its dependencies installed via Cabal. Once GHC and Cabal are installed, you can navigate to the src directory and run the build-ws.sh and build-er.sh shell scripts to install dependencies and respectively build the WebSocket and CLI interfaces.

Using the GUI

After successfully running build-ws.sh you can now run the local WebSocket server via bin/ws-server, which will listen on localhost:4243.

While the WebSocket server is running, simply open ui/app.html in your browser (no need for an HTTP server, it will work fine when opened via file: URI). You should now be able to configure and run simulations. The Save and Load buttons can store one configuration in browser Local Storage.

Using the CLI

You can execute bin/exp-runner to more efficiently perform simulations based on a JSON config file. It will output a CSV file with periodic propagation results. Refer to the experiments directory for example configurations.

You can also use the GUI to create configuration files: whenever you click Start, ws-server will output the JSON corresponding to the current configuration to stdout. This can be stored in a file and directly fed to exp-runner.

Currently available models

The following mobility, identification and infection models are currently implemented and available via the GUI and CLI:

• Mobility models:
  • Lévy Walk model by by Rhee et al.
  • BSA model by Haas
  • SMOOTH model by Munjal et al.
• Identification models:
  • Dummy: all agents are immediately identified by all others
  • Signal: signal profile model described in the paper
  • Gaussian, PeriodicTrigger and Broadcast: experimental; replaced by Signal model
• Infection models:
  • Serial: The serial infection model employed by Zyba et al.
  • Simple: experimental parallel infection model with failure probabilities.

Implementing a new component model

You can implement a new component model by adding a Haskell module that provides an implementation of the SimulationModel type class, as well as an implementation of either MobilityModel, IdentificationModel or InfectionModel. The data type implementing these classes contains the parameters of your model. It should also derive FromJSON so that it can be dynamically configured.

To enable support in the CLI, edit the module Model.UI.JSON and add the model to one of the lists mobModels, idModels or infModels. You can provide a name here that is used to identify the model in the JSON configuration file.

Refer to the existing model classes in src/Model for examples.