ICT Research

An Effective Statistical Model to Evaluate the Two-phase Query Algorithm in the Encrypted Database with A Secure Index Structure

Ninh Duong-Bao — 2025-10-11

Database security is essential for protecting sensitive data in the database from unauthorized access, manipulation, or theft. To access the encrypted data, the two-phase query algorithm controls and filters database queries to ensure that only authorized data is accessible. Traditional methods used qualitative analyses to evaluate the effectiveness of the algorithm and were unable to completely judge the degree of credibility. In this paper, a statistical model is proposed that provides quantitative analyses to evaluate the effectiveness of the two-phase query algorithm in the encrypted database with a secure index structure using three major statistical quantities, including completeness, effectiveness, and credibility degrees. From the experimental results conducted in various databases, the proposed model is proven to be efficient and suitable for evaluating the two-phase query algorithm. Moreover, the stability of the statistical quantities is also discussed in different query conditions.

Integrating Explainable AI for Transparent and Accurate Cryptocurrency Prediction

Anh Hoang — 2025-10-11

The cryptocurrency market, particularly Bitcoin, is highly volatile, making accurate price prediction a crucial challenge in finance and machine learning. This study employs the Extreme Gradient Boosting algorithm to develop a robust predictive model for Bitcoin price prediction. Using historical price data and technical indicators, the model captures intricate market patterns and enhances predictive accuracy. Our results indicate that XGBoost achieves a strong R-squared score of 0.9267, demonstrating high reliability for short-term forecasting. The model is trained on high-frequency historical data, incorporating features such as moving averages, the relative strength index, and Bollinger bands. To improve interpretability, we integrate Explainable AI techniques, specifically SHapley Additive ExPlanations and Local Interpretable Model-agnostic Explanations. SHAP highlights that recent closing prices significantly influence predictions, while LIME provides instance-specific explanations. By combining SHAP’s global interpretability with LIME’s localized insights, we bridge the gap between predictive accuracy and transparency. We also address potential risks such as overfitting and data leakage through rigorous cross-validation and hyperparameter tuning. Our findings suggest that XGBoost, when properly optimized and interpreted, serves as a powerful tool for short-term cryptocurrency price forecasting. This study contributes to AI-driven financial analytics, offering valuable insights for traders, analysts, and researchers seeking data-driven decision making in cryptocurrency markets.

ENHANCING HATE SPEECH DETECTION WITH KNOWLEDGE-BASED PROMPTING FOR LARGE LANGUAGE MODELS

Thanh Le — 2025-10-11

The widespread growth of social media has intensified the challenge of identifying and addressing hate speech, largely due to its intricate semantics and varied contextual nuances. This study examines the potential of employing prompting strategies with large language models (LLMs) alongside conventional deep neural networks (DNNs) and pre-trained language models for hate speech detection. The models assessed include DNNs such as Text CNN and GRU, PLMs like BERT, XLM-R, DistilBERT, and PhoBERT, as well as LLMs leveraging prompting techniques, including GPT. Experiments conducted on both Vietnamese and English datasets show that GPT, utilizing the Chain-of-Thought reasoning framework, achieves the best macro F1 scores—particularly when enhanced with a knowledge-based prompting (KBPrompt) approach. In KBPrompt, domain-specific knowledge bases consisting of slang and emoji terms are constructed directly from the training data using prompt-assisted extraction and a frequency-based labeling strategy, which assigns labels based on their empirical distribution across annotated comments. The findings highlight the effectiveness of incorporating external lexical knowledge into LLM prompting pipelines, especially in multilingual and low-resource contexts, to better capture implicit toxicity and culturally embedded hate speech.