Enhancing Cybersecurity through Ransomware Detection using Hybridization of Heuristic Feature Selection with Deep Representation Learning Model

Maha Farouk

doi:https://doi.org/10.54216/JCIM.170204

Journal of Intelligent Systems and Internet of Things | 2026

Hybrid Neural Networks and Machine Learning for Detection of Diabetic Retinopathy

Diabetic retinopathy (DR) is one of the most common causes of blindness in the world, and early detection plays an important role in therapy. In this paper, we introduce a hybrid framework with the merger of sophisticated image processing techniques and deep learning models for automated DR detection from retinal fundus images. Information starts with an extensive preprocessing pipeline, which includes bilateral filtering for noise reduction, removal of artifacts, adaptive contrast enhancement and a precise segmentation in the U-Net architecture. To increase model robustness, random rotation augmentation was used to mimic different imaging positions. GLCM analysis is used to extract texture features capturing important lesion-related patterns, and deep features are extracted using a fine-tuned EfficientNet-B0 model. The hybrid feature set is then modelled by a Support Vector Machine (SVM) with the radial basis function kernel and optimized with cross-validation and hyperactive parameters. Experiments show our model can well solve the image heterogeneity problem and yields a high level of accuracy in diagnosis and grading corresponding severity requirements of DR stage.

groups

Waleed Khalid Al-zubaidi mail -

Shokhan M. Al-Barzinji mail -

Zaid Sami Mohsen mail -

Omar Muthanna Khudhur mail

link https://doi.org/10.54216/JISIoT.180213

Volume & Issue

Vol. Volume 18 / Iss. Issue 2

Details open_in_new

Journal of Intelligent Systems and Internet of Things | 2026

Residual Graph Convolutional Networks for Improving Rumor Detection from Social Media Texts

Internet and social media have become significant platforms for sharing real-time information, with rumors significantly affecting billions of people's perceptions. Considerably, Rumor recognition is the most challenging task on social media platforms. Numerous Deep Learning (DL) models have been developed to extract linguistic characteristics from short-text tweets for rumor prediction. However, these models struggles to capture the intricate spatiotemporal relationships presenting tweet interactions. To address this issues, Bidirectional Encoder Representation from Transformers with Attention based Balanced Spatial-Temporal Graph Convolutional Networks (BERT-ABSTGCN) was used. This model incorporates Spatial-Temporal Attention Mechanism (STAM) and a Spatial-Temporal Convolution Module (STCM) to effectively model the spatiotemporal dependencies within in tweet interactions to enhance rumor detection.  However, it constitutes to high degradation problem due to convergence issues. A popular solution to these problems is Residual Learning (RL), which introduces identity mappings to speed up training and enhance gradient propagation. However, traditional RL can only be used for layer-wise task refining, which severely restricts its capacity to grasp more generalized dependencies. However, conventional RL is restricted to layer-wise refinement within a single task limiting its ability to capture broader dependencies. To address this, the proposed work is included with a Cross-Residual Learning (CRL) in BERT-ABSTGCN named BERT with Attention-based Balanced Spatial-Temporal Residual Graph Convolutional Networks (BERT-ABSTRGCN) for efficient rumor detection and stance classification. CRL of BERT-ABSTRGCN enable intuitive learning across multiple tasks like rumor detection and stance classification using cross-connections. CRL establishes direct connections between shallow and deep feature representations, mitigating the vanishing gradient issue.   The fitted residual mappings in the CRL will facilitate the BERT- BERT-ABSTRGCN with the provided information by using the short cut connections and lowers the probability of model degradation. BERT-ABSTRGCN effectively identifies rumor with different stances about specific social media posts, thereby preventing the spread of rumors. Experimental evaluations show that BERT-ABSTRGCN achieves 95.62% accuracy on the PHEME dataset and 90.15% on Mendeley’s COVID-19 rumor dataset, significantly surpassing traditional models.

groups

Vanitha Siddheswaran mail -

Prabahari Raju mail

link https://doi.org/10.54216/JISIoT.180215

Volume & Issue

Vol. Volume 18 / Iss. Issue 2

Details open_in_new

Journal of Intelligent Systems and Internet of Things | 2026

Deep Fake Image Detection Using Ensemble Approach

This paper offers a comprehensive framework for real or fake image classification based on three classifiers: a Standard Convolutional Neural Network (CNN), an EfficientNetV2 model based on transfer learning, and a re-trained GAN discriminator to address the challenges in deepfake detection. The CNN, with four convolutional blocks and dropout regularization, offers computational efficiency (87.2% accuracy, 15 ms/image inference), while EfficientNetV2 utilizes pre-trained ImageNet weights to achieve state-of-the-art performance (94.7% ac-curacy, AUC: 0.98) using hierarchical feature extraction. The fine-tuned and adversarial-pretrained GAN discriminator demonstrates niche strength in the detection of synthetic artifacts (91% recall for GAN-generated fakes). Training used augmented sets (rotation, shifts, and shear) to increase the generalization boost, with loss optimization and early stopping (binary cross-entropy) controlled through validation. Normalized test set validation affirmed EfficientNetV2's capability at balancing recall (94%) with precision (95%), although the GAN discriminator recorded a lead in adversarial resilience. All the models blended, an ensemble model achieved maximum accuracy (96.1%), under complementarities. Computational baselines showed trade-offs EfficientNetV2 accu-racy vs. resource bias (2.5-hour training), the CNN edge-compatibility, and the GAN discriminator arti-fact-sensitive specialization. The work encourages hybrid architectures and ensemble approaches to balance out single-model vulnerabilities, offering a flexible toolkit for deepfake warfare while emphasizing the need for hardware-aware deployment techniques and ongoing adaptation to changing synthetic approaches.

groups

Vijay Madaan mail -

Raghad Tohmas Esfandiyar mail -

Shahad Hussein Jasim mail -

Oday Ali Hassen mail -

Neha Sharma mail -

Ansam A. Abdulhussein mail

link https://doi.org/10.54216/JISIoT.180214

Volume & Issue

Vol. Volume 18 / Iss. Issue 2

Details open_in_new

International Journal of Neutrosophic Science | 2025

Neutrosophic Midrange Measure in Bayesian Selection

Many of the problems that we face in our lives and daily work are how to directly and accurately select candidates or categories from multiple sets of candidates (categories). The ranking and selection approach is a modern and direct method for selecting categories easily, which is associated with a probability of correct selection. In this paper, we employ the neutrosophic Bayes procedure for decision to select multinomial population. Select a mid-range category for multiple categories and employ neutrosophic logic to define a modern Bayesian procedure that incorporates parameters with some indeterminacy and has a prior distribution, which we call the neutrosophic prior distribution.

groups

Kawther F. Alhasan mail

link https://doi.org/10.54216/IJNS.260420

Volume & Issue

Vol. Volume 26 / Iss. Issue 4

Details open_in_new

International Journal of Neutrosophic Science | 2026

Neutrosophic Average Edge Connectivity with Applications to Communication Networks

Average edge connectivity is a fundamental concept in graph theory, widely employed to evaluate the robustness of networks through the analysis of local edge cuts. Classical fuzzy extensions allow for graded membership, yet they fail to clearly distinguish between inherent uncertainty and definite absence of edges. To overcome this limitation, we introduce the notion of neutrosophic average edge connectivity, a tri-valued connectivity measure formulated within the framework of single-valued neutrosophic graphs (SVNGs). In this study, we rigorously define neutrosophic local edge cuts, establish key theoretical results including bounds and monotonicity properties, and design efficient algorithms tailored for particular families of graphs. The applicability of the proposed framework is demonstrated through a detailed communication-network case study, which highlights its capacity to capture structural resilience under indeterminate conditions. Overall, the proposed approach generalizes classical robustness indicators and provides a comprehensive tool for analyzing connectivity in networks characterized by vagueness, indeterminacy, and incomplete information.

groups

Aparna Tripathy mail -

Amaresh Chandra Panda mail -

Siva Prasad Behera mail -

Prasanta Kumar Raut mail -

Mana Donganont mail -

Said Broumi mail

link https://doi.org/10.54216/IJNS.270214

Volume & Issue

Vol. Volume 27 / Iss. Issue 2

Details open_in_new

International Journal of Neutrosophic Science | 2025

Coefficient Bounds for Generalized n-Fold Symmetric Neutrosophic Bi-univalent Functions

In this paper, we introduce and investigate new generalized subclasses of neutrosophic n-fold symmetric bi-univalent functions defined in the open unit disk U . These subclasses are characterized via four neutrosophic multi-parameters κ, ρ, γ, and β, which provide a flexible framework to capture the truth, indeterminacy, and falsity components inherent in geometric and analytic behaviors. Within this neutrosophic setting, we derive upper bounds for the initial coefficients |dn+1| and |d2n+1|, and establish generalized Fekete–Szeg˝o inequalities for the considered classes. The results obtained extend and unify several existing results in classical and neutrosophic bi-univalent function theory. Examples and corollaries are presented to demonstrate the sharpness and applicability of the results.

groups

Isra Al-Shbeil mail -

wael mahmoud mohammad salameh mail -

Jianhua Gong mail -

Ajmal Khan mail -

Shahid Khan mail

link https://doi.org/10.54216/IJNS.260419

Volume & Issue

Vol. Volume 26 / Iss. Issue 4

Details open_in_new

Journal of Cybersecurity and Information Management | 2026

A Hybrid Intelligent Facial Recognition Model Based on Hierarchical Feature Extraction and Il-lamination Normalization

Face recognition in unconstrained environments is difficult due to varying poses and lighting conditions. This can severely impair the performance of intelligent recognition models. Traditional methods often do not adapt well to these variations, which results in poor performance and limited applicability. This paper proposes a hybrid intelligent face recognition model based on hierarchical feature extraction and illumination normalization (H-FR). The proposed method employs a hierarchical feature extraction model to capture macro and micro facial details, ensuring reliable recognition across diverse poses and lighting conditions. Employing Adaptive Histogram Equalization on the A and B channels of the LAB colour space effectively normalizes illumination variations, enhancing the visibility and consistency of facial features. The proposed model has been tested and validated on the "Pins Face Recognition" dataset available on Kaggle, which encompasses various celebrity faces captured in varying poses and lighting conditions. The proposed model has been demonstrated through extensive experimentation to outperform AlexNet and VGG-19. The compared algorithms achieved accuracies of 88% for AlexNet and 93% for VGG-19, while the proposed H-FR model achieved 96%.

groups

Ali F. Rashid mail -

Ilyas Khudhair Yalwi mail -

Ali Hakem Alsaeedi mail -

Riyadh Rahef Nuiaa Alogaili mail -

Mazin Abed Mohammed mail

link https://doi.org/10.54216/JCIM.170201

Volume & Issue

Vol. Volume 17 / Iss. Issue 2

Details open_in_new

Journal of Cybersecurity and Information Management | 2026

MACSteg: Real-Time Voice Authentication and Deepfake Protection Using Device MAC Address Steganography

The invention of deepfake applications make it possible to produce highly natural and real voice recordings which creates critical concerns about the credibility of audio telecommunications. The confirmation of the speakers’ voices became essential especially for sensitive data such as financial, healthcare, and surveillance risk management services, authentication of speakers’ voices became significantly crucial. To improve solutions to this issue, this paper presents MACSteg strategy which is a real-time, lightweight voice authentication technique by discreetly encapsulate device’s MAC address within voice file using Quantization Index Modulation (QIM) stego-technique. Unlike many traditional strategies that degrade voice quality or produces noticed jitter, MACSteg technique preserve both clarity and efficiency. Implementations showed that the hidden MAC address stayed intact in spite of some typical voice processing such as compression, while interfered signals reformed by clatter or volume variations were consistently detected. The proposed system obtained a high signal-to-noise ratio (SNR) exceeding 70 dB, illustrating that the alterations were inaudible, and maintained well in real-time submissions, giving only a processing delay of 0.01 milliseconds per each audio segment. The results indicate MACSteg’s potential as a ascendable and effective approach for safeguarding voice authenticity, especially in circumstances where verification of speaker’s voice is vital.

groups

Sanaa Ahmed Kadhim mail -

Zaid Ali Alsarray mail -

Saad Abdual Azize Abdual Rahman mail -

Massila Kamalrudin mail -

Mustafa Musa mail

link https://doi.org/10.54216/JCIM.170202

Volume & Issue

Vol. Volume 17 / Iss. Issue 2

Details open_in_new

Journal of Cybersecurity and Information Management | 2026

Integration of Crayfish Optimization with Watermarking Scheme for Automated Tampering Text Detection

Digital text document serves as the essence of existing communication but still pose major safety concerns on their vulnerability to tampering. Digital text watermarking acts as a powerful tool to secure the reliability of textual data. Presenting a hidden layer of safety and accountability enables organizations and individuals to make sure of the truth behind each file and trust the written word. Watermarking detects tampering by checking the embedded signature for changes or distortions. The Watermark model is capable of mechanically repairing and classifying themselves once tampered with, enhancing document resilience. Watermarking is an effective mechanism to identify tampering attacks in digital documents. The specialized process of embedding imperceptible and strong watermarks in document creation or distribution detects alterations. This study proposes the Crayfish Optimization with a Watermarking Scheme for Automated Tampering Text Detection (CFOWS-ATTD) technique. The major purpose of the CFOWS-ATTD technique is to accomplish the security of English text using content authentication and tampering detection. In the CFOWS-ATTD technique, two-stage processes are involved. Moreover, the CFOWS-ATTD technique generates a watermark from the text document and performs extraction to verify text authenticity. Furthermore, the CFO approach optimally places the watermark to ensure it remains robust and imperceptible to tampering. The experimentation of the CFOWS-ATTD approach is performed under the ELST, ESST, EHMST, and EMST datasets. The results implied that the CFOWS-ATTD approach obtains optimum performance over other techniques.

groups

Hanan T. Halawani mail

link https://doi.org/10.54216/JCIM.170203

Volume & Issue

Vol. Volume 17 / Iss. Issue 2

Details open_in_new

Journal of Cybersecurity and Information Management | 2026

Enhancing Cybersecurity through Ransomware Detection using Hybridization of Heuristic Feature Selection with Deep Representation Learning Model

Network security has become vulnerable to hacker threats owing to its advancement and easily accessible to computer and internet technology. Ransomware is the most commonly used malware in cyberattacks to mislead the victim user to expose private and sensitive data to hackers. Ransomware is malicious software that encodes the entire system or consumer’s files, creating it impossible, and later demands a payment fee from the victim’s computer in exchange for the decryption key. Ransomware attacks become highly popular and overwhelming for both individuals and organizations. Recently, deep learning (DL) and machine learning (ML) models are established to identify ransomware attacks in real-time and categorize them into various types. The system will be considered to examine the behaviors of malicious software and detect the particular kind of ransomware being utilized. This data will enhance the system’s accuracy and deliver appropriate data to cybersecurity professionals and victims. Therefore, this study proposes an accurate Ransomware Detection and classification using the Hybrid Metaheuristic Feature Selection with Deep Learning (RDC-HMFSDL) technique. The aim is in effectually detecting and classifying the ransomware attacks. Initially, the RDC-HMFSDL technique utilizes min-max model to transform the input data into a standard setting. Furthermore, the hybrid red deer sparrow search optimization (HRDSO) approach is used for the feature selection (FS). For ransomware attack detection, the long short-term memory autoencoder (LSTM-AE) approach is employed. Finally, the sine cosine algorithm (SCA) is used to optimally choose the parameter values of the LSTM-AE approach. The RDC-HMFSDL approach was tested on a benchmark dataset, achieving a superior accuracy of 99.88% compared to existing methods.

groups

Maha Farouk Sabir mail

link https://doi.org/10.54216/JCIM.170204

Volume & Issue

Vol. Volume 17 / Iss. Issue 2

Details open_in_new