Journal on Information Technologies & Communications

Automatic temperature monitoring system for pigs using ear tag sensors integrated with bluetooth and machine learning

Thanh Nam Duong — 2025-12-20

The automatic temperature monitoring system for pigs, integrated with ear tag sensors combined with Bluetooth Low Energy (BLE) technology and machine learning models, offers a modern and efficient solution for monitoring the health of pig herds. The sensors are directly attached to ear tags and are capable of continuously measuring each pig’s body temperature, while also collecting environmental information such as humidity and ambient temperature inside the housing facility. Data is transmitted to a central server via BLE and analysed on a cloud computing platform. The system employs advanced machine learning models such as PCA-SVC, PCA-Decision Tree, PCARandom Forest, and 1D-CNN to enable early detection of abnormal signs. Among these models, 1D-CNN demonstrates outstanding performance, achieving an accuracy of 96.58%, proving its effectiveness in processing time-series data and identifying health issues. The application of smart sensor technology not only enhances monitoring accuracy and reduces labor costs but also optimises real-time surveillance capabilities. This contributes significantly to disease prevention and improves overall livestock management efficiency.

Research on control takeover techniques for counter-UAV (Unmanned Aircraft Vehicle)

Cong Thanh Bui*, Duc Dung Phung, Xuan Chinh Dong — 2025-12-20

Unmanned aerial vehicles (UAVs) have become increasingly integral in both civilian and military domains due to their flexibility, cost-effectiveness, and mission endurance. However, this widespread adoption introduces complex security challenges, especially in terms of managing and preventing unauthorized UAV activities. Among the emerging counter-UAV techniques, hijacking UAVs by exploiting control protocol vulnerabilities is considered a promising and cost-efficient approach, particularly for research and defense applications. This paper presents technical findings related to hijacking-based countermeasures against UAVs by targeting the MAVLink protocol - a widely adopted lightweight communication protocol for UAV command and telemetry. The research outlines the architecture of UAV control systems, identifies protocol vulnerabilities and demonstrates how spoofed control packets can lead to full UAV takeover. The results show that, due to the lack of encryption and authentication in MAVLink 1.0, an attacker can successfully disrupt and override mission commands, effectively hijacking UAVs. This virtual simulation provides a valuable resource for education, training, and further research on UAV cybersecurity, particularly in environments where access to real UAV systems and field conditions is limited. The study emphasizes the importance of protocol security in UAV systems and opens new directions for proactive defense strategies.

Detecting software vulnerabilities using deep learning

Van Cong Bui — 2025-12-20

Developing software projects successful is always a top concern for organizations and enterprises. Among these concerns, ensuring software quality is the highest priority throughout the entire development and operation process. This paper addresses the detection of source code vulnerabilities and focuses on analyzing the syntax and semantics of statements within the source code. The source code vulnerability detection model follows a structured process: (i) syntactic and semantic representation; (ii) feature extraction from source code; (iii) data balancing; and (iv) source code classification. The model’s output indicates whether the source code is normal or contains vulnerabilities. The model is trained using the SARD dataset and incorporates deep learning approaches. Specifically, it employs the BERT (Bidirectional Encoder Representations from Transformers) model, the Word2Vec model combined with LSTM (Long Short-Term Memory), and the Word2Vec model with BiLSTM (Bidirectional Long Short-Term Memory) across three scenarios. Classification results are passed through a softmax function to generate a vector containing the probability predictions for each type of vulnerability. The detection model achieves an accuracy rate of up to 82.63% for identifying source code vulnerabilities, with a corresponding omission rate of only 17.37%. This result is considered acceptable and demonstrates the superior effectiveness of the approach in the task of source code vulnerability detection.

Application of jitter augmentation in classification of Raman scattering from glucose liquid with several machine learning models

Thanh Tung Nguyen — 2025-12-20

The utilisation of machine learning technology for invasive diabetes diagnosis has emerged as a prominent research area in the medical field in recent years. The study employs Raman spectroscopy of a glucose fluid sample to ascertain glucose levels. We produce glucose-liquid samples with 14 varying ratios of glucose to pure water to replicate the 14 glucose concentrations found in human blood and acquire the corresponding Raman spectra. With the optimistic result of the previous study, we also design the jittering augmentation approach to enhance the dataset and the dataset size was increased by a factor of 100. We employ a 1-D Convolutional Neural Network to ascertain glucose levels in samples. The result proves the positive effect of jitter augmentation for improving the performance of the deep learning model as well as high accuracy of these models in classification of Raman scattering.

Adaptive buffer strategy for regular High-Utility sequential pattern Mining on incremental data

Minh Thai Tran — 2025-12-20

Regular High-Utility Sequential Pattern Mining (RHUSP) from incremental databases poses a significant challenge in balancing computational efficiency and accuracy. Current algorithms using a fixed buffer ratio (µ) often suffer from limitations: a low µ value results in wasted memory resources, while a high µ carries the risk of missing important patterns. This paper proposes Adaptive-RIncHusp, an algorithm featuring an adaptive buffer strategy that automatically adjusts the parameter µ for each data batch update. The method integrates three complementary heuristics: utility Ratio, Threshold difficulty, and growth rate, combined with a high-sensitivity risk function (f x)=x^0.25) for early detection of data fluctuations. Experiments on 5 benchmark datasets demonstrate that Adaptive-RIncHusp is 8-15% faster and saves 20-40% memory compared to the conservative fixed strategy (µ=0.4) while maintaining recall > 97%. Furthermore, the algorithm achieves speeds comparable to the aggressive fixed strategy (µ=0.9) while completely eliminating pattern loss, confirming its effectiveness in volatile data environments.

Efficient mining of succinct correlated high utility association rules

Thong Tran — 2025-12-20

This paper investigates the problem of mining high utility association rules from quantitative databases. Traditional approaches mainly rely on conventional utility-confidence measures without considering the actual strength of association among items. Due to the absence of a correlation measure, many discovered rules exhibit high utility while the items within the rules co-occur infrequently, making them difficult to interpret and of limited practical significance; moreover, the number of generated rules is often very large. To address these limitations, this paper proposes a novel utility-confidence measure and integrates the Bond correlation measure to ensure that the items in a rule not only yield high utility but also exhibit strong correlations. On this basis, the paper introduces a class of succinct correlated high utility association rules, called SCoHUARs, in which each rule is represented in a compact form that serves as a representative of an equivalence class of rules sharing the same support. In addition, a partitioning method for the set of SCoHUARs is presented to guarantee complete and non-redundant generation of all SCoHUARs, together with several optimization techniques to accelerate the mining process. Finally, the M-SCoHUAR algorithm is designed to efficiently mine SCoHUARs by integrating the proposed theoretical results. Experimental results demonstrate that the proposed algorithm significantly outperforms existing methods in terms of execution time, memory consumption, and the quality of the discovered patterns.