In this chapter, a comprehensive pipeline for detecting vulnerabilities in critical infrastructures is presented, focusing on the classification of SCADA images into IT and OT categories using deep learning. The study leverages transfer learning and fine-tuning on a custom dataset (CRINF-300) to evaluate multiple CNN architectures, comparing their performance in terms of accuracy, F1-score, and computational efficiency, and discusses experimental findings and future directions.
(UPI) Chapter 13: Detecting Vulnerabilities in Critical Infrastructures Course (How To)
via Treehouse
Overview
Syllabus
13.3 Methodology
In this stage, the chapter outlines the methodology by reviewing the state of the art in CNN-based image classification, detailing the construction of the CRINF-300 dataset, and proposing a pipeline that employs transfer learning and fine-tuning to classify ICS screenshots.
Chevron 5 steps- instruction
13.2. State of the Art
-
Quiz 13.2
5 questions
- instruction
13.3.1 Critical Infrastructure Dataset
- instruction
13.3.2 Proposed Pipeline
-
Quiz 13.3
5 questions
13.4. Experimental Results and Discussion
In this stage, the experimental settings, results, and comparisons of various CNN architectures are discussed, highlighting key performance metrics and the trade-offs between speed and accuracy in both transfer learning and fine-tuning approaches.
Chevron 3 steps- instruction
13.4.1 Experimental Settings
- instruction
13.4.2 Discussion of Results
-
Quiz 13.4
5 questions
