Trace2TAP: Synthesizing Trigger-Action Programs From Traces of Behavior

Introduction

This repository reproduces the result from our paper "Trace2TAP: Synthesizing Trigger-Action Programs From Traces of Behavior" based on anonymized data collected from our user study.

You can build the fullstack web-application we used in our user study which are executed in several docker containers.

Pre-requisites

You need to install docker, docker-compose and git-lhs on your local machine.

• Setting up docker
• Setting up docker-compose
• Setting up git-lhs

The following building process does not support Windows or Windows Subsystem Linux. We tested it on native Ubuntu 18.04 and 20.04.

Download Trace2TAP

git clone https://github.com/zlfben/trace2tap.git

Build Trace2TAP

./start_server_dev.sh --data data/data-anonymous.sql data/iotcore.sql

This will build all our images and run the containers. It will start a local web-application at port 80.

Test our interface

Goto "localhost/" in the browser and use the following credentials to log in with our participants trace (anonymized).

• Username: superifttt-p* (p=1-7)
• Password: anything (We disabled the authentication process for local run).

See the results in our user study

Interview coding

The coding of the interview can be found at data/Trace2TAP-Coding.xlsx.

Rules synthesized in the interview

In the terminal, run:

docker-compose -f docker-compose.dev.yml exec backend python3 manage.py review list

This shows a list of pages participants saw in their interviews. To check one of them in detail, run:

docker-compose -f docker-compose.dev.yml exec backend python3 manage.py review show -i <id>