Performance Analysis of Explainable Deep Learning-Based Intrusion Detection Systems for IoT Networks: A Systematic...

Area of Science:

• Cybersecurity
• Artificial Intelligence
• Internet of Things

Background:

• Deep learning (DL) models in IoT intrusion detection systems (IDS) lack transparency, impacting trust and reliability.
• Explainable AI (XAI) aims to improve interpretability but its effect on performance in resource-constrained IoT is unclear.

Purpose of the Study:

• To systematically review the performance trade-offs of explainable DL-based IDSs for IoT networks.
• To analyze detection accuracy, computational overhead, and explanation quality.
• To identify gaps and propose solutions for practical deployment.

Main Methods:

• Systematic literature review following PRISMA methodology.
• Analysis of 129 peer-reviewed studies (2018-2025).
• Investigation of XAI technique trade-offs, DL architectures, and deployment challenges.

Main Results:

• Existing approaches often achieve high detection accuracy at the cost of computational efficiency and explainability.
• This imbalance limits the practical deployment of explainable IDSs on IoT edge devices.
• Significant gaps exist in post-deployment XAI evaluation practices.

Conclusions:

• A unified framework is needed to model the trilemma between performance, efficiency, and explainability in IoT IDSs.
• Proposed XAI evaluation framework standardizes post-deployment metrics.
• Actionable insights are provided for developing trustworthy and efficient explainable IDS for IoT.