Active vs. Passive? A Comparison of Automata Learning Paradigms for Network Protocols

This repository contains the supplemental material to the paper ' Active vs. Passive? A Comparison of Automata Learning Paradigms for Network Protocols' of Bernhard K. Aichernig¹, Edi Muškardin^1,2, and Andrea Pferscher¹ (¹Institute of Software Technology, Graz University of Technology; ² Silicon Austria Labs, TU Graz - SAL DES Lab).

Content

• BLE models (automata/BLE/):
  • CYBLE-416045-02 [PDF]
  • nRF52832 [PDF]
  • CC2650 [PDF]
  • CYW43455 [PDF]
  • CC2640R2 (no pairing request) [PDF]
  • CC2640R2 (no feature request) [PDF]
  • CC2652R1 [PDF]
• MQTT models (automata/MQTT/):
  • ActiveMQ [PDF]
  • emqtt [PDF]
  • HBMQTT [PDF]
  • Mosquitto [PDF]
  • VerneMQ [PDF]
• Experiment execution (main.py):

Prerequisites

Python library Aalpy >=1.2.8

Requirements installation:

sudo pip3 install -r requirements.txt

Experiment Execution

python3 main.py <BENCHMARK>

Possible benchmarks are "MQTT" or "BLE". The script generates three different csv-files containing the learning results for active learning, passive learning, and active learning with cache.

Acknowledgement

• AALpy: active automata learning library