Explainability for Property Violations in Cyber-Physical Systems: An Immune-inspired Approach

Overview

This repository contains the code implementation for the paper titled "Explainability for Property Violations in Cyber-Physical Systems: An Immune-inspired Approach." The goal of the project is to provide an immune-inspired approach to enhance the explainability of property violations in cyber-physical systems.

Table of Contents

• Abstract
• Getting Started
  • Prerequisites
  • Installation
• Usage
• Contact

Abstract

Complex relations between cybernetic and physical components of a cyber-physical system (CPS) in tandem with continuous environment changes represent a challenge to engineering robust CPSs. To help engineers determine the cause of violations, there is a need for a systematic approach that helps understanding the system behaviors that lead to critical failures of the CPS. In this work, we present a methodology that identifies and isolates crucial anomalous behaviors that can not only hamper the system but also are often challenging to capture while engineering a CPS.

Getting Started

Prerequisites

$ pip install -r requirements.txt

Usage

Following the paper's structure, the code was split into files inside src/ and numbered for convenience.

Before the method

• 1.app_create_experiment_markov.py: responsible for collecting the patient's vital signs based on the markov chain developed in this Github repo. A csv with the readings is stored for the next steps.

Step 1: Feature Engineering

• 2.data_pipeline.py: Runs the simulations in OpenModelica based on the readings. The results are parsed and the features are created.
• 3. Temperature_sanity_check.ipynb: Checks the sanity of the engineered features.
• 4. temperature_eda.ipynb: Explores the dataset with the engineered features. Features are ranked based on the correlation with the label and low correlated ones are discarded.

Step 2: Negative Selection

• 5. Negative Selection.ipynb: Execution of the NSA algorithm to generate the detectors

Step 3: Detector's Analysis

• 6. temperature_ns_result.ipynb: Analysis of the detectors, generation of the clusters and plots.

Contact

If there are any questions regarding the overall methodology or how to run the experiments, please feel free to contact us at:

• jp.araujo@hu-berlin.de