Reliable evaluation for the AI-enabled intrusion detection system from data perspective.

Area of Science:

• Cybersecurity
• Artificial Intelligence
• Data Science

Background:

• Intrusion detection systems (IDS) are crucial for cybersecurity.
• Existing AI models for IDS overlook the impact of poor data quality.
• Data integrity issues like tampering and poisoning erode trust in AI-driven IDS.

Purpose of the Study:

• To propose a multi-indicator comprehensive evaluation method (MICEM) for assessing intrusion detection data quality.
• To enhance the reliability and usability of AI-enabled IDS by addressing data quality concerns.
• To provide a data-centric approach for validating AI decision-making in cybersecurity.

Main Methods:

• Established several evaluation indicators to analyze potential risks in intrusion detection data.
• Developed specific quantitative methods for assessing data quality dimensions.
• Conducted a comprehensive evaluation using MICEM to determine overall data quality.

Main Results:

• The proposed MICEM effectively evaluates intrusion detection data quality.
• Quantitative indicators were developed to identify data risks.
• Comprehensive evaluation guarantees the reliability of AI-enabled IDS.

Conclusions:

• MICEM ensures the reliability of AI decision-making in IDS by focusing on data quality.
• The method addresses data integrity issues, mitigating trust crises in AI models.
• Effectiveness and practicality were validated on benchmark and real-world datasets.