International Journal of Autonomous and Adaptive Communications Systems (16 papers in press)

Regular Issues

• Cyber Security Automation and Managing Cyber Threats in Network through Smart Techniques: An Intelligent Approach for Future Gen. Systems  
  by Rohit Rastogi, Vaibhav Sharma, Tushar Gupta, Vaibhav Gupta 
  Abstract: Cybersecurity has become a major concern in this digital era. Since, the cyberattacks and their types are increasing at an immense rate, it is not humanly possible to monitor, identify and take actions against the attacks. With the current automation systems majorly relying on supervised learning algorithms where they have already seen the type of attacks to monitor and manage the attacks, these systems have been rendered inefficient by zero day attacks. The immense potential of AI and utilise it to its full potential in the field of cybersecurity. If correctly applied, Artificial Intelligence can help to detect and deal with the cyberattacks more efficiently and can help protect users that are not very security conscious and are not aware about the dangers of these security breaches. The authors have decided to utilise machine learning algorithms like decision trees and knowledge discovery in database (KDD) to detect zero day attacks as well as handle other common cyberattacks.
  Keywords: Supervised Learning; Unsupervised Learning; KDD (Knowledge Discovery in Database); phishing; smashing; DDoS.
  DOI: 10.1504/IJAACS.2025.10063962
   
  
• DDoS attack detection and prevention model using Pipit Flying Fox optimization-based Deep Neural network  
  by Anuja Sharma, Parul Saxena 
  Abstract: The software-defined network (SDN) remains the futuristic model that helps to satisfy the new application demands of future networks. However, the control panel of SDN is the prime target of destructive attacks, especially distributed denial of service (DDoS). The restrictions in the conventional techniques such as reliability to network topology, low accuracy, and hardware dependencies manifest the need for effective DDoS detection. Hence, the research develops a DDoS attack recognition and prevention model aid with an optimised deep learning network. The significance relies on the pipit flying fox (PPF) optimisation, which selects the optimal hyperparameters, minimises the errors, and accelerates the learning speed. The experimental results are reported as the specificity, sensitivity, and accuracy of 98.5551%, 92.4951%, and 98.4951% respectively for 80% of training. Further, the values are obtained as 98.6397%, 86.0997%, and 98.09972% for specificity, sensitivity, and accuracy respectively at K-fold 10 which exceeds other competent techniques.
  Keywords: SDN; DDoS attack; security; attack detection; Deep learning; optimization.
  DOI: 10.1504/IJAACS.2025.10064035
   
  
• Joint 5G NR Polar Code-Convolutional Code design for Massive MIMO-UFMC system  
  by Smita Jolania, Ravi Sindal, Ankit Saxena 
  Abstract: Polar codes (PC) are the major contender in fifth generation-New Radio (5G-NR) for error control in the physical downlink control channel (PDCCH) The work proposes a novel concatenated error correction technique of PC with convolutional codes (CC) and is experimented under 5G simulation constraints. This research paper develops a simulation model of Universal Filtered Multicarrier (UFMC) modulation based massive multiple-input multiple output (MIMO) technique targeting for short burst transmissions. The UFMC uses sub-band filtering with reduced out of band emission (OOBE) and enhanced spectral efficiency. An analytical framework of the novel PC-CC-UFMC system to effectively correlate the flexible design parameters for different wireless channels is implemented to enhance Bit Error Rate (BER) performance. The results shown in paper, a gain in the required Signal to Noise Ratio (SNR) for same BER is reduced by approximately 5dB for increase in antenna from 64 to 256.
  Keywords: Polar codes; New Radio; convolutional codes; Massive MIMO; UFMC.
  DOI: 10.1504/IJAACS.2025.10064049
   
  
• Deepfake Detection Based on Single-Domain Data Augmentation  
  by Qian Feng, Zhifeng Xu 
  Abstract: Deepfake has posed a serious threat to personal privacy and social stability The related research on deepfake detection has gained sufficient high accuracy on various datasets, while the generalisation performance is still insufficient Most of the existing methods are aimed at analysing and detecting specific traces and distortions generated by a specific forgery algorithm. However, these detection algorithms typically experience a significant decline in accuracy when detecting forgery videos generated by other algorithms This paper proposed a Deepfake detection scheme based on Single-Domain Data Augmentation, and considered the most difficult situation in the deepfake detection generalization problem: How to generalise to a variety of unknown forgery data when only the real data is known We proposed the Universal Forgery Generation (UFG) and Adversarial Style transfer algorithm (AST) to augment forgery data and improve generalisation ability The experimental results show that our scheme is superior to many existing schemes.
  Keywords: Deepfake detection; Domain generalisation; Style transfer.
  DOI: 10.1504/IJAACS.2025.10064478
   
  
• ADMET Property Prediction Model Based on Feature Selection and Data Mining Techniques  
  by Gu Junlin, Xu Yihan, Sun Juan, Liu Weiwei 
  Abstract: Breast cancer has posed a significant threat to women's health in recent years, and the search for compounds that can antagonize ER? activity will play an important role in breast cancer treatment. ADMET properties are important indicators of compound efficacy, and existing research has used machine learning techniques to fit collected data, but with some performance limitations. In this paper, we use data mining techniques to establish a biologically active-ADMET property prediction model. Firstly, important features were obtained through feature selection techniques, and 23 feature variables that have an impact on ADMET properties were selected. Then, LightGBM and genetic algorithms were used for biological activity prediction tasks, and the R2 value on the validation set reached 0.75, achieving good performance. Finally, based on the BP neural network, the ADMET-UMLP model was constructed, proposing a U-shaped structure to fully utilize the underlying feature information. The model performed well on the validation set, with AUC values exceeding 0.9 in the classification prediction of Caco-2, CYP3A4, hERG, HOB, andMNproperties, and a prediction of 0.98 AUC value for Caco-2, demonstrating good predictive performance.
  Keywords: ADMET; LightGBM; machine learning; prediction.
  DOI: 10.1504/IJAACS.2025.10064499
   
  
• Dual-scale Dual-rate Video Compressive Sensing for Edge Surveillance Device  
  by Yue Lu, Zhang Xiang, Chengsheng Yuan 
  Abstract: Classic video compression method suffers from long encode time and requires large memories, making it hard to deploy on edge devices, thus video compressive sensing which requires less resources during encoding is getting more attention. We propose a dual-scale dual-rate video compressive sensing algorithm for surveillance video compression. Proposed method extracts and compresses foreground area and reference frame separately using dual-scale compressive sampling, then using reversible neural network to reconstruct original frames. Finally we test compressive sampling and ROI extraction network in proposed method on edge device and reconstruction network on server. The experiments show that proposed method can fast compresses frame and extracts foreground area on edge computing devices, achieves higher reconstruction quality.
  Keywords: Video compressive sensing; reversible neural network; surveillance video; siamese network; edge computing; neural processing unit; RK3399 Pro.
  DOI: 10.1504/IJAACS.2025.10064730
   
  
• Analysis of secrecy performance under double shadowed -  
  by Damepaia Lato, Rajkishur Mudoi 
  Abstract: In this paper, the physical layer security (PLS) under the double shadowed - fading channel is investigated. Being a composite fading channel model, it is a realistic representation of the propagation environment in which wireless signals experience a complex interplay of different phenomena. The mathematical statements of the secrecy outage probability (SOP) and the probability of non-zero secrecy capacity (PNSC) have been investigated by considering one legitimate receiver and one eavesdropper listening to the source transmitting confidential information. Based on the obtained mathematical expressions, the secrecy performance metrics are analysed and the results are plotted for both the SOP and the PNSC. It can be observed that as the signal-to-noise ratio (SNR) of the legitimate user increases, the SOP reduces and the secrecy capacity increases for the double shadowed - distributions.
  Keywords: Composite fading channel; Double shadowed ?-? fading channel; Secrecy capacity; Physical layer security; Secrecy outage probability.
  DOI: 10.1504/IJAACS.2025.10064742
   
  
• A Deep Neural Network for Fashion Retrieval Based on Multi-Attention Attribute Manipulation  
  by Qianyi Liu, Jiaohua Qin, Xuyu Xiang, Yun Tan 
  Abstract: The surge in online shopping has heightened the demand for interactive fashion design retrieval. Existing methods, however, exhibit imperfections in attribute segmentation, attributed to the specificity of clothing attributes. The attention region often encounters multiple attributes overlapping, causing changes in one attribute to affect irrelevant ones, resulting in poor retrieval accuracy. This paper addresses this challenge by proposing a deep neural network for fashion retrieval based on multi-attention attribute manipulation. In this approach, the feature extraction module sifts the extracted features to obtain an overall description of the clothing image by adding ESE-NAM combined attention modules to the VoVNet network block. The attribute decoding module utilizes one-hot coding and feature mapping to subdivide the attribute features, obtaining more independent local detail features for refined attribute image retrieval with a focus on details. Experimental results show that the proposed network surpasses exsiting networks with an overall accuracy increase of more than 4 percentage points, particularly with the feature extraction module demonstrating an accuracy boost of over 6 percentage points.
  Keywords: image retrieval; fashion design retrieval; interactive image retrieval; deep neural network; deep learning.
  DOI: 10.1504/IJAACS.2025.10065081
   
  
• Enhancing Malayalam Question Classification in Question Answering Systems: A Comparative Study of SVM, KNN, and Multinomial NB  
  by Bibin P. A, Ravisekhar R, Babu Anto P 
  Abstract: The method of question classification, involving the analysis and assignment of questions to specific categories, has gained momentum due to increased online activity, prompting interest in automating this process into predefined categories. The study focuses on developing a machine learning based model for classifying question types in a Malayalam Question Answering System (QAS). It begins with systematic preprocessing of the dataset and feature extraction, followed by partitioning into training and testing sets. Three machine learning algorithms including support vector machine (SVM), multinomial Naive Bayes (MNB), and K-nearest neighbour (KNN) are implemented and optimised using various hyper-parameters. The evaluation employs metrics like accuracy, precision, recall, F1-score, and confusion matrices to assess performance comprehensively. Results indicate that the SVM classifier achieves the highest accuracy among the models tested. The research underscores the effectiveness of machine learning techniques in automating question classification, especially in diverse linguistic contexts like Malayalam, facilitating more efficient question-answering systems.
  Keywords: Malayalam Question Answering; Multinomial Naïve Bayes; SVM; KNN; Question classification; Machine learning; TF-IDF; n-gram.
  DOI: 10.1504/IJAACS.2025.10065484
   
  
• Deep Learning based PSA Detection Model in Multi-user M-MIMO Networks  
  by Manju V.M, Ganesh R.S 
  Abstract: It is commonly known that MASSIVE MIMO (M-MIMO) is a key component for the forthcoming wireless networks. Base stations (BSs) in M-MIMO networks are fitted with an enormous number of antennas to provide several advantages over conventional MIMO, including easier power control, improved spectrum efficiency, and increased efficiency of energy. Since the estimated CSI might be contaminated by the eavesdrop interaction; M-MIMO systems are susceptible to pilot spoofing attacks (PSAs), which result in significant information leakage in the downstream transmission. To protectgenuine communications, this work introduces a new PSA detection model in multiuser M-MIMO (MU M-MIMO). Initially, signal transmission takes place and then the large scale fading factors are estimated. Further, PSA detection is done using Deep Neural Network (DNN) framework. Finally, the optimal channel estimation is done using Self Customized Black Widow Optimization Algorithm (SC-BWO). Moreover, analysis is performed on error probability, BER and so on.
  Keywords: M-MIMO; Pilot spoofing attacks; Multiuser; Deep Neural Network; SC-BWO Algorithm.
  DOI: 10.1504/IJAACS.2025.10067442
   
  
• MASCNN: Speech Emotion Recognition using Multi-head Area Self-Attention Convolutional Neural Network  
  by Qianli Ma, Wenjie Zhang, Yan Diqun 
  Abstract: Speech Emotion Recognition (SER) is pivotal for enhancing human-computer interaction by interpreting emotional expressions. Traditional machine learning approaches often encounter limitations in accuracy and adaptability. This study introduces a novel deep learning-based SER method incorporating attention mechanisms. Data augmentation techniques, including noise injection and speech modification, are applied prior to feature extraction. The Mel-frequency Cepstral Coefficient (MFCC) features are then processed using a Convolutional Neural Network (CNN) enhanced with a Multi-head Area Self-attention mechanism to improve emotion classification. Evaluation on the IEMOCAP database demonstrates that the proposed method outperforms existing SER techniques in recognition accuracy, with data augmentation significantly enhancing model performance. Additionally, guided class speech elicitation proves more effective than deductive class speech.
  Keywords: Speech Emotion Recognition; Attention Mechanism; CNN; IEMOCAP.
  DOI: 10.1504/IJAACS.2025.10068924
   
  
• Privacy-Preserving Outsourced Library Resource Data Sharing Scheme Based on Proxy Re-Encryption  
  by Tianzhang Li, Cai Mei, ChunXiao Zhang 
  Abstract: In the current digital era, libraries are increasingly incorporating cloud-based solutions to manage extensive collections and provide users with access to diverse resources. However, the adoption of cloud computing introduces critical challenges, particularly in ensuring data confidentiality and protecting user privacy. This paper proposes an innovative privacy-preserving library data sharing method based on proxy re-encryption, specifically designed for cloud environments. Our scheme enables secure outsourcing of library resources to the cloud, ensuring that only authorised users can access these resources. A key innovation of our approach is its ability to conceal users search queries from the cloud server, safeguarding query privacy. Moreover, library administrators can dynamically update encrypted data stored in the cloud without compromising security. Compared to existing methods, our work introduces a novel access control mechanism that enhances security while significantly reducingcommunicationcomplexity,requiringonlyoneroundofcommunication half the amount of prior approaches. Through comprehensive security analysis and performance evaluation, we highlight the distinct advantages of our method in maintaining data confidentiality, integrity, and operational efficiency
  Keywords: Privacy-preserving; Library resource; Proxy re-encryption; Outsourced.
  DOI: 10.1504/IJAACS.2025.10070012
   
  
• Software Security Assurance with an Augmented Software Component Analysis Approach for Open Source Component Evaluation  
  by Jian Hu, Linfei Li, Hailin Wang, Tao Chuan, Xiwei Dai, Yaodan Yu, Jie Wang 
  Abstract: Open source components are the foundation of modern software development, thereby making Software Component Analysis (SCA) as an essential method to ensuring software security. However, existing SCA methods concentrate on identifying open source component's security issues, ignoring comprehensive analysis with components' maintenance and support posture. In this paper, we investigated the literature from industry and academia about secure software development process, software and component evaluation models, and evaluation methodologies. Then, we proposed AOSCA, an augmented SCA approach by additionally quantifying a set of attributes for evaluation, despite of the conventional detected issues. The experimental results demonstrate that AOSCA can effectively assessing open source components and provide the evaluation result to software development organization based on requirements and preference. To sum up, AOSCA provides a comprehensive but effective mechanism for open source component evaluation. Applying AOSCA as a security practice during software development process is vital for software security assurance.
  Keywords: software security; software component analysis; open source component; security evaluation; augmented attribute.
  DOI: 10.1504/IJAACS.2025.10070268
   
  
• STGN: Spatio-Temporal Graph Network For Few-shot Cross-Domain Image Steganalysis  
  by Mingqian Liu, Daqiu Li, Zhang Xiang, Zhangjie Fu 
  Abstract: In the actual image steganalysis task, it is difficult for steganalysis models to obtain large-scale steganographic images of unknown steganography as training datasets. Inspired by few-shot learning, we propose a novel spatio- temporal graph network (STGN) for few-shot cross-domain steganalysis. Firstly, we design a multi-domain feature preprocessing network in the spatial, frequency, and feature domains, so that STGN can extract the deep steganographic features from different domains. Secondly, we design a spatio-temporal graph convolution network is designed to extract the effective spatio-temporal components in the steganalysis sequence feature; Then, the spatio-temporal steganalysis feature is as the input of graph network to classify. Finally, the STGN is trained on the benchmark datasets including the steganographic images of spatial and frequency domains.Through intra- domain and cross-domain testing, experimental results show that the average accuracy is above 85.55% (1-shot) and 93.97%(5-shot).
  Keywords: image steganalysis; few-shot learning; cross-domain steganalysis; deep Learning.
  DOI: 10.1504/IJAACS.2025.10071616
   
  
• 5G-Powered Digital Twin for Orchestration and Prediction Task based on Cascaded LSTM  
  by Sridharan Kannan, Ahmad Y. A. Bani Ahmad, Thella Preethi Priyanka, Yogapriya J 
  Abstract: An advanced deep learning-based 5G-powered DT replica for orchestration and monitoring purposes is implemented in this paper. The major aim of this paper is the development of the MNDT replication, which is capable of being implemented in a 5G core environment. This model is found helpful in various advanced applications like Industry 4.0 and cyber security. The required model for the development of the MNDT is gathered by utilising a data acquisition approach. Further, the modelling is performed by making use of bidirectional connections within the digital and physical elements. Once the data collection is complete, and the MNDT is developed, then the second phase of the work is executed. Here, the prediction is carried out using the Cascaded Long Short-Term Memory (CasLSTM) framework. Experimental validations are carried out to prove the efficacy of the implemented deep learning-based 5G-powered MNDT for the prediction model.
  Keywords: 5G Technology; Digital Twin; Cascaded Long Short Term Memory; Data Acquisition; Orchestration and Prediction Task.
  DOI: 10.1504/IJAACS.2025.10071683
   
  
• A Robust Generative Coverless Co-steganography Method of Text and Image  
  by Yuxi Deng, Yun Tan, Le Mao, Xuyu Xiang 
  Abstract: Most existing coverless image steganography methods based on image generation suffer from insufficient hiding capacity. To address this, a Generative Coverless co-steganography method of text and image is proposed. Secret text is first converted into noise vectors using convolutional error-correcting codes, then a container image is generated using a self-attention generative adversarial network (SAGAN). The secret image is converted into a binary sequence through pixel encoding, and a style image containing the secret information is generated by an encoder. These images are then input into the generator to produce two sets of generative stego images for transmission. The secret text is recovered using a gradient descent method based on the simulated annealing algorithm, while the secret image is recovered through the decoder. Evaluated on CelebA and Fashion MNIST datasets, compared to existing methods, the proposed method significantly improves capacity and efficiency, increasing hiding capacity by 1.6 times.
  Keywords: Coverless steganography; Co-steganography of text and image; Generative adversarial networks.
  DOI: 10.1504/IJAACS.2025.10071727