Journal of ICT Research and Applications

Enhancing Natural Language Inference Performance with Knowledge Graph for COVID-19 Automated Fact-Checking in Indonesian Language

Arief Purnama Muharram — 2025-09-15

Automated fact-checking is a key strategy to overcome the spread of COVID-19 misinformation on the internet. These systems typically leverage deep learning approaches through natural language inference (NLI) to verify the truthfulness of information based on supporting evidence. However, one challenge that arises in deep learning is performance stagnation due to a lack of knowledge during training. This study proposes using a knowledge graph (KG) as external knowledge to enhance NLI performance for automated COVID-19 fact-checking in the Indonesian language. The proposed model architecture comprises three modules: a fact module, an NLI module, and a classifier module. The fact module processes information from the KG, while the NLI module handles semantic relationships between the given premise and hypothesis. The representation vectors from both modules are concatenated and fed into the classifier module to produce the final result. The model was trained using the generated Indonesian COVID-19 fact-checking dataset and the COVID-19 KG Bahasa Indonesia. Our study demonstrates that incorporating KGs can significantly improve NLI performance in fact-checking, achieving a maximum accuracy of 0.8616. This suggests that KGs are a valuable component for enhancing NLI performance in automated fact-checking.

AI-enhanced Cybersecurity Risk Assessment with Multi-Fuzzy Inference

Essam Natsheh — 2025-09-15

The pace and complexity of modern cyber-attacks expose the limits of traditional ‘impact × likelihood’ risk matrices, which compress uncertainty into coarse categories and miss inter-dependent threat dynamics. We propose a three-layer multi-fuzzy inference system (MFIS) that models general infrastructure vulnerabilities and access-control weaknesses separately, then fuses them into a single, continuous 0-25 risk score. The framework was validated on three representative scenarios—catastrophic/continuous, serious/frequent, and minor/few attacks—encompassing sixteen threat criteria. Compared with a crisp 5 × 5 matrix, MFIS cut mean-absolute error and root-mean-square error by 90 to 99% and reproduced expert-panel judgments to within 0.55 points across all scenarios. Nine independent practitioners rated the prototype highly on usability (100% agreement), credibility (100%) and actionability (100%), with 78% willing to recommend adoption. These results demonstrate that MFIS delivers fine-grained, expert-aligned assessments without adding operational complexity, making it a viable drop-in replacement for time- or resource-constrained organizations. By capturing partial memberships and cross-domain interactions, MFIS offers a more faithful, adaptive and explainable basis for prioritizing cyber-defense investments and can be extended to emerging threat domains with modest rule-base updates.