International Journal of Internet Protocol Technology (9 papers in press)

Regular Issues

• Concept Drift Detection Method for Noisy Big Data Streams based on Least Squares Line Fitting  
  by Zhihong Qin, Zhanlei Shang, Fuhong Geng 
  Abstract: In order to improve the accuracy of concept drift detection in big data streams, a noisy big data streams concept drift detection method based on least squares line fitting is proposed. Firstly, the Kalman filter model is used to identify and remove noise components from the original data stream, and then the wavelet thresholding algorithm is used to filter out the remaining noise. Secondly, by fitting the denoised data stream with a least squares line, linear trend features are extracted to simulate the trend of concept drift. Finally, the Tr-OEM algorithm is used to identify concept drift in big data streams, and an efficient and accurate concept drift detection is achieved through a machine learning model that dynamically adapts to data changes. The experimental results show that the detection accuracy of our method remains between 96.89% and 97.45%, and the maximum average change distance does not exceed 1.
  Keywords: Least squares linear fitting; Noisy big data stream; Concept drift detection; Kalman filtering.
  DOI: 10.1504/IJIPT.2025.10072667
   
  
• Effective Information Extraction of Massive IoT Big Data based on Adaptive Spectral Clustering Algorithm  
  by Lin Yu, Shumin Zhi 
  Abstract: To improve the standardised mutual information (NMI) and average accuracy (AA) of effective information extraction from IoT big data, a method for extracting effective information from massive IoT big data based on adaptive spectral clustering algorithm is proposed. Firstly, build a multi-layered IoT platform that utilises multiple sensors to collect and process multimodal data. Secondly, based on spatial discreteness, feature partitioning and differential evolution are performed on multi-type data models, and SVM algorithm is used for dimensionality reduction. Combining the aggregation tree frequent pattern set to analyse the correlation dimension, filtering node interference and extracting effective information features of networked big data. Finally, by using adaptive spectral clustering algorithm, combined with adaptive parameters and nearest neighbour path distance, effective information from networked big data can be accurately extracted. Experimental data shows that the NMI values of this method are all above 0.8, and the AA values are above 90%.
  Keywords: spectral clustering; Internet of Things; Massive data; Effective information extraction.
  DOI: 10.1504/IJIPT.2025.10072668
   
  
• Secure Sharing of Electronic Medical Records based on Blockchain and Searchable Encryption  
  by Hongliang Tian, Aomen Zhao 
  Abstract: To solve the problems of long response time, low throughput, and high risk of data loss in current electronic medical record security sharing methods, a secure sharing method of electronic medical records based on blockchain and searchable encryption is proposed. The electronic medical record sharing architecture is built using blockchain technology. In this architecture, symmetric searchable encryption technology is introduced, and a parallel search mechanism is used to efficiently retrieve encrypted electronic medical record data while protecting data privacy, thus achieving secure sharing of electronic medical records. The experimental results show that this method has shorter response time and higher throughput in high concurrency situations, with a maximum response time of only 500ms and throughput ranging from 2340TPS to 2830TPS. Moreover, the data loss is extremely low in malicious attack environments, providing strong support for the development of smart healthcare
  Keywords: Electronic medical records; Secure sharing; Blockchain; Searchable encryption; Data privacy; Parallel search mechanism.
  DOI: 10.1504/IJIPT.2025.10072669
   
  
• Faultload Sequences for the MQTT Protocol Services  
  by Amina Jandoubi, M.Taha Bennani, Olfa Mosbahi 
  Abstract: We have proposed preliminary results of a new approach to extract the sequences of events that model the fault injection times at the MQTT messaging protocol level to assess their reliability. After targeting publish service, we got seven possible scenarios to put the system in a faulty state. We introduced new algorithms to: (1) extract send and receive events from CFGs, (2) identify send/receive pairs, (3) time stamp the events using a vector clock, (4) filter sending events, and (5) generate alternative send sequences. In this paper, we extended the initial results by applying the approach to the services provided by the Publisher, the Subscriber, and the MQTT Broker, which increased the number of scenarios up to seventeen fault injection sequences. This contribution is the first step in formalising the description of all possible attacks on the MQTT protocol, which is crucial to reinforce the reliability of its services.
  Keywords: MQTT; Internet of things; Fault injection; Control-Flow Graphs.
  DOI: 10.1504/IJIPT.2026.10072721
   
  
• An Optimisation Method for Intelligent Legal Information Query based on Association Rule Data Mining Algorithm  
  by Haimei Yu 
  Abstract: Aiming to address the issues of low accuracy, low recall, and long average response time in current intelligent legal information query optimization methods, a new optimisation method for intelligent legal information query based on association rule data mining algorithm is proposed. Optimise the association rule data mining algorithm based on ant colony algorithm, obtain legal information using the optimized association rule data mining algorithm, and synchronise and fill in missing legal information obtained. Build an intelligent legal information query optimisation model based on the processed legal information, and use an improved genetic algorithm to solve the model, thereby achieving intelligent legal information query optimisation. The experimental results show that the maximum accuracy of the proposed method for intelligent legal information query is 98.86%, the maximum recall rate is 97.21%, and the average response time varies between 0.21s and 0.65s, demonstrating high precision and efficiency.
  Keywords: Association rule data mining algorithm; Legal information; Optimization method for intelligent legal; Ant colony algorithm; Improved genetic algorithm.
  DOI: 10.1504/IJIPT.2026.10072850
   
  
• Risk Assessment of Network Public Opinion Information Circle based on CAFA-BP Algorithm  
  by Qianqian Zheng 
  Abstract: To address the issues of low coverage of influencing factors, low evaluation accuracy, and long response time in current risk assessment methods for network public opinion information circle dissemination, a risk assessment method of network public opinion information circle based on CAFA-BP algorithm is proposed. Optimise the LSTM model through attention mechanism, use the optimized LSTM model to predict the popularity of network public opinion, and determine the factors affecting communication risk based on the popularity of network public opinion. Using CAFA algorithm to optimise BP neural network, establish a risk assessment model based on CAFA-BP algorithm, and use this model to determine the risk value of network public opinion information circle propagation. The experimental results show that the maximum coverage rate of the influencing factors of the proposed method is 98.56%,the evaluation accuracy varies between 94.58% and 97.62%, and the response time varies between 1.23s and 1.69s.
  Keywords: CAFA-BP algorithm; Network public opinion information circle; Risk assessment; Popularity of network public opinion.
  DOI: 10.1504/IJIPT.2025.10072909
   
  
• Corporate Financial Risk Prediction Model Based on Deep Learning  
  by Fei Xue 
  Abstract: The correlation between financial success and corporate social responsibility (CSR) has been the subject of extensive research in recent years While prior studies have explored this relationship, objectively assessing the effectiveness of CSR programs has remained a considerable challenge However, advancements in research methodologies and the development of Environmental, Social, and Governance (ESG) measurement criteria have facilitated more reliable evaluations With the rapid advancement of artificial intelligence, and deep learning (DL) techniques have been increasingly applied across various domains The application of machine learning to ESG data analysis remains relatively limited Consequently, deep learning methodologies have proven to be practical, efficient, and effective approaches for predicting corporate financial performance based on ESG metrics Through risk management optimization, accuracy and recall rates have improved to 0 905 and 0 890, respectively The proposed risk prediction model achieves high accuracy rates of 90 5% and 92 4% in the training and test data sets, respectively.
  Keywords: Key performance indicator (KPI); Sustainable Development Goal (SDG); Return on equity (ROE); spiking neural networks (SNNs); artificial neural network (ANN); Deep learning (DL); machine learning (ML).
  DOI: 10.1504/IJIPT.2025.10072912
   
  
• Optimising AdaBoost with Bio-Inspired Algorithms for Stock Market Prediction: a Comparative Study on Hang Seng Index  
  by Cuicui Feng, Kai Huang, Wendi Wang, Xian Yu 
  Abstract: This study presented a new integrated forecasting model for improving the accuracy of stock price predictions for the Hang Seng Index. Using the ensemble learning theory, this research constructs the core modelling engine based on Adaptive Boosting (AdaBoost) is fine-tuned using three metaheuristic optimisation approaches: Particle Swarm Optimisation, Antli-on Optimisation, and Biogeography-Based Optimization. These algorithms are based on nature search for optimizing Ada-Boost hyperparameters such as the number of estimators, learning rate, and complexity of the base learner to improve generalisation and minimize prediction errors. The BBO-AdaBoost was the most superior one. A dataset including daily HSI stock prices- open, high, low, close, and volume-were obtained from 2015 to 2023. The data were min-max normalized and were split into 80:20 for training and testing. The results show that the BBO-AdaBoost achieved the best result with the least MAPE of 2.29%, outperforming both the stand-alone AdaBoost and other hybrid models.
  Keywords: Stock Price Prediction; Hang Seng Index; Adaptive Boosting; Biogeography-Based Optimisation; Particle Swarm Optimisation; Antlion Optimisation; Hybrid Forecasting Model; Machine Learning; Financial M.
  DOI: 10.1504/IJIPT.2026.10072913
   
  
• Effective Information Classification Mining Method for IoT Big Data Based on Support Vector Machine  
  by Shumin Zhi 
  Abstract: The effective information classification mining of IoT big data is of great significance in enhancing data value, optimising resource allocation, and promoting technological innovation to solve the problems of low recall and precision, and long completion time in current methods. Research on effective information classification mining method for IoT big data based on support vector machine has been carried out. By optimising the selection of initial k values and initial clustering centre points, the improved K-means clustering algorithm is utilized to achieve big data collection in the Internet of Things. By predicting residuals for anomaly detection and removal of IoT data, the removed IoT big data is input into a fuzzy support vector machine to achieve information classification mining. The experimental results show that the average recall rate of the proposed method is 97.07%,the average precision rate is 96.72%, and the task completion time varies between 0.22s and 0.68s.
  Keywords: Support vector machine; IoT; Big data; Effective information; Classification mining; Improved K-means clustering algorithm.
  DOI: 10.1504/IJIPT.2025.10072914