Description

This repository contains a discrete event simulator to evaluate trust-based interactions between agents with limited memory.

Memory is represented via four buffers with fixed sizes:

1. Cryptographic
2. Trust
3. Reputation
4. Stereotypes

Different buffer eviction strategies are implemented to evaluate which produces the highest utility of information contained in those buffers.

Requirements

sudo apt-get install python3 cm-super graphviz
python3 -m pip install matplotlib numpy scipy hmmlearn tqdm pygraphviz seaborn

Running an Experiment

run.sh defines running multiple eviction strategies and agent behaviours for different sized buffers. run_multiple.py can be used to obtain a large number of repeats and run-graphics.sh can be used to obtain graph results for an individual run.

python3 run_multiple.py

To perform a run with a specific seed call run.sh or run-graphics.sh such as:

SEED=10 ./run.sh

Analysing Results

Graphing Individual Results

./analyse_individual.py create graphs for a single simulation.

./graph_individual.py <Behaviour>/<Eviction Strategy>/<buffer size>-metrics.<seed>.pickle.bz2

Combining Results

Individual results need to be combined for them to be graphed

./combine_results.py VeryGoodBehaviour/* AlwaysGoodBehaviour/* UnstableBehaviour/* GoodBehaviour/*

Graphing Combined Results

./graph_multiple.py VeryGoodBehaviour/* AlwaysGoodBehaviour/* UnstableBehaviour/* GoodBehaviour/*

Creating a graph showing space usage

Trust models with different sizes will consume different amounts of space. The following script can be used to visualise this.

./graph_space.py VeryGoodBehaviour/* AlwaysGoodBehaviour/* UnstableBehaviour/* GoodBehaviour/*

Creating graphs of buffer evolution over time

The following script can be used to visualise the evolution of buffer content over time.

mkdir -p "<out-dir>/"
./graph_buffers.py <Behaviour>/<Eviction Strategy>/<buffer size>-metrics.<seed>.pickle.bz2 --path-prefix "<out-dir>/"