Potential and limitations of machine learning modeling for forecasting Acute Food Insecurity

Mélissande Machefer

doi:10.1016/j.gfs.2025.100859

Interpretable and robust intrusion detection: A hybrid CBAM-CNN model with XAI techniques

2026-06-20

0

PRE

AI

S

Saheed Ademola Bello

F

Farid Binbeshr *

K

Khalid Ibrahimi

M

M. Waleed Shinwari

M

Muhammad Imam

A

Ashraf Mahmoud

DOI：10.1016/j.compeleceng.2026.111348

Original

Original request for help

Save

Abstract

En 中文

• Innovative Hybrid Architecture: Combines a 1D-CNN with a Convolutional Block Attention Module (CBAM) to automatically focus on the most critical network features while remaining lightweight. • High Detection Performance: Achieves an exceptional 99% accuracy on standard datasets and remains robust (85% accuracy) even against modern, complex IoT-based cyberattacks. • Human-Readable Explanations: Uses XAI techniques (SHAP and LIME) to transform complex AI decisions into clear, actionable insights that help security analysts understand exactly why a threat was flagged. • Real-Time Ready: Optimized for speed with a very low latency (3.8 ms), making it practical for immediate deployment in live Security Information and Event Management (SIEM) systems.

Interpretable and robust intrusion detection: A hybrid CBAM-CNN model with XAI techniques

Abstract

Journal

Organization

Cited Papers

Citing Papers