Abdullah Ghanim Jaber

Mr. Abdullah Ghanim Jaber

Higher Education, University of Information Technology and Communications, Baghdad, Baghdād, Iraq
+1
orcid

Research Focus & Metrics

Academic Career Length

5Years (2022-Present)

(Calculated once Scilit publication data is synced: current year − year of first publication + 1)

Representative Publications

View All Publications
31 October 2022 in Applied Sciences
Verified
Journal Article
Open access
PDF file
Abdullah Ghanim JaberAbdullah Ghanim Jaber
Abdullah Ghanim Jaber
Ravie Chandren MuniyandiRavie Chandren Muniyandi
Ravie Chandren Muniyandi
LATEEF USMANLATEEF USMAN
LATEEF USMAN
HSHarprith Kaur Rajinder Singh
Harprith Kaur Rajinder Singh
Face recognition has grown in popularity due to the ease with which most recognition systems can find and recognize human faces in images and videos. However, the accuracy of the face recognition system is critical in ascertaining the success of a person’s identification. A lack of sufficiently large training datasets is one of the significant challenges that limit the accuracy of face recognition systems. Meanwhile, machine learning (ML) algorithms, particularly those used for image-based face recognition, require large training data samples to achieve a high degree of face recognition accuracy. Based on the above challenge, this research proposes a method for improving face recognition precision and accuracy by employing a hybrid approach of the Gabor filter and a stacked sparse autoencoders (SSAE) deep neural network. The face image datasets from Olivetti Research Laboratory (OLR) and the Extended Yale-B databases were used to evaluate the proposed hybrid model’s performance. All face image datasets used in our experiments are grayscale image type with a resolution of 92 × 112 for the OLR database and a resolution 192 × 168 for the Extended Yale-B database. Our experimental results showed that the proposed method improved face recognition accuracy by approximately 100% for the two databases used at a significantly reduced feature extraction time compared to the current state-of-art face recognition methods for all test cases. The SSAE approach can explore large and complex datasets with minimal computation time. In addition, the algorithm minimizes the false acceptance rate and improves recognition accuracy. This implies that the proposed method is promising and has the potential to enhance the performance of face recognition systems.
1

12 March 2026 in International Journal of Research and Innovation in Applied Science
Verified
Journal Article
Open access
PDF file
Abdullah Ghanim JaberAbdullah Ghanim Jaber
Abdullah Ghanim Jaber
The paper presents a new framework, HE-FedStress, with which it is feasible to monitor stress and preserve privacy in an AI-based approach to stress detection through federated learning and homomorphic encryption on wearable edge devices. The proposed system handles the essential privacy issues that come with decentralized health surveillance because the proposed system allows joint model training without exposing uncoded biometric information. The Edge wearable devices are local physiological signal temporal attention models that inference latency can process five-second windows with a latency of just 47 ms, implemented on photoplethysmography, electrodermal activity and accelerator data. The updates of the model are coded using the Pailier cryptosystem and sent to the central server where they could be aggregated securely without decrypting each individual contribution. The WESAD and SWELL-KW datasets have been empirically assessed to verify that HE-FedStress can give F1-scores of 89.7% and 85.2% respectively and retain centralized model performance with 92 and 94 % of this performance under full cryptographic protection against model-inversion attacks (compared to 34.7 % centralization in the standard federated learning case). The structure uses gradient quantization, which cuts down the communication load to 63KB/ round, and uses adaptive batch size to support heterogeneous device capacity. The computational optimizations (depthwise separable convolutions, selective attention pruning, and eight-bit quantization) allow incessant 24 hour execution of commercial wearables with only a 19% power consumption impact over plaintext federated learning. The design of the modular architecture is to meet GDPR and HIPAA data-localization standards and provide the ability to extend it to other healthcare applications. Therefore, this contribution can provide a practical, scalable stress monitoring solution to secure and personalized performance with a robust guarantee on cryptography, and can also indicate that the requirements of resource-constrained edge environments can be satisfied with a robust guarantee alongside real-time performance.

18 November 2025 in Preprints.org
Verified
Preprint
Open access
PDF file
Abdullah Ghanim JaberAbdullah Ghanim Jaber
Abdullah Ghanim Jaber
Ravie Chandren MuniyandiRavie Chandren Muniyandi
Ravie Chandren Muniyandi
KAKhairul Akram Zainol Ariffin
Khairul Akram Zainol Ariffin
The exponential rise of remote sensing and sensor network technologies has generated massive amounts of visual data, posing challenges in safe transmission, data integrity, and categorization. In real-time applications, existing approaches fail to reconcile strong encryption, classification accuracy, and computing economy. This study aims to develop a safe and effective remote sensing image classification system that addresses both data security and intelligent analysis, enabling automated, context-aware insights in dispersed, real-time settings. The proposed work introduces the Fox-Optimized Secure Hybrid Image Encryption and Learning-based Detection (FOX-SHIELD) framework, which effectively integrates advanced encryption techniques with deep learning-based image classification, ensuring both data security and high classification accuracy for remote sensing images in real-time, distributed environments. An upgraded Chebyshev chaotic map and the Secure Hash Algorithm (SHA-256) provide dynamic, stable encryption keys in the first phase, ensuring data secrecy and integrity throughout transmission and storage. A Fast Recurrent Neural Network (FRNN) coupled with the Fox Optimization Algorithm improves convergence rate, stability, and classification accuracy even for encrypted input in the second phase. This integration enables powerful object detection while ensuring anonymity, an essential feature for sensitive remote sensing tasks. The FOX-SHIELD framework outperforms traditional models, including Convolutional Neural Networks (CNNs), Recurrent Neural Networks (RNNs), and other hybrid encryption-learning models, in terms of classification accuracy, training convergence, and computational efficiency when applied to standard remote sensing datasets. This work addresses the fundamental issue of data security in remote sensing image classification by integrating lightweight cryptographic methods with metaheuristic deep learning optimization to enhance model accuracy, convergence, and computational efficiency in real-time applications.

22 March 2026 in Dijlah Journal of Engineering Sciences ISSN: 3078-9664, e-ISSN: 3078-9656
Verified
Journal Article
Open access
PDF file
Abdullah Ghanim JaberAbdullah Ghanim Jaber
Abdullah Ghanim Jaber
AAAbeer Ahmed Ali
Abeer Ahmed Ali
AMAli A. Mahmood
Ali A. Mahmood
MSMohammed Jamal Salim
Mohammed Jamal Salim
GMGhaith Jaafar Mohammed
Ghaith Jaafar Mohammed
We suggest a new method of enhancing the feature discriminability of mammogram analysis through integrating anatomical pretext tasks in a hybrid CNN-ViT backbone followed by improved classification using SVM. Traditional self-supervised approaches usually use generic image transforms, which does not necessarily reflect the subtle clinical reasoning regarding radiologists. To tackle this, we formulate domain-specific pretext tasks, which directly model anatomical priors, such as spatial context reconstruction, orientation prediction conditional on ductal tree alignment and lesion-context consistency using contrastive learning. The hybrid backbone is a hybrid of ResNet-50 and ViT, which extracts local and global context patterns, respectively, to produce an integrated feature representation that links hierarchical and long-range dependencies. These characteristics are then trimmed down using PCA, so that they can be compatible with SVM kernels whilst retaining anatomical significance. In comparison to the current approaches, our framework is the only one to use the pretraining goals with clinical workflows, thus enhancing the interpretability of features and minimizing the need of large labeled datasets. Experiments indicate that the suggested approach performs better than the conventional deep feature extractors in mammogram classification tasks. Integrating domain-sensitive medical image analysis with self-directed learning through incorporating anatomical reasoning provides new opportunities to analyze medical images, especially in cases when few annotations are available. The contribution of the paper is the progression of the synergy between radiologist-inspired feature learning and up-to-date deep architectures, which gives a scalable algorithm that enhances the accuracy of diagnostic outcomes.

1 March 2026 in Al-Iraqia Journal for Scientific Engineering Research
Verified
Journal Article
Open access
PDF file
Abdullah Ghanim JaberAbdullah Ghanim Jaber
Abdullah Ghanim Jaber
AAAbeer Ahmed Ali
Abeer Ahmed Ali
AMAli A. Mahmood
Ali A. Mahmood
MSMohammed Jamal Salim
Mohammed Jamal Salim
GMGhaith Jaafar Mohammed
Ghaith Jaafar Mohammed
KAKhairul Akram Zainol Ariffin
Khairul Akram Zainol Ariffin
This paper presents a dynamic quantization-aware neural architecture search (NAS) framework for real-time encrypted traffic classification in 5G networks. The framework jointly optimizes model architecture and quantization policies using reinforcement learning with hardware-in-the-loop feedback, addressing both accuracy and efficiency on edge devices. Unlike conventional static methods, the proposed system supports runtime bitwidth switching to adapt precision levels based on fluctuating traffic loads and hardware conditions. Extensive experiments were conducted on ISCX-VPN, USTC-TFC2016, and QUIC-5G datasets. The results show that the proposed approach achieves 94.2% accuracy on ISCX-VPN, 92.7% on USTC, and 89.4% on QUIC-5G, outperforming baseline methods while reducing inference latency by 28–42% and lowering energy consumption by up to 26%. The framework maintains robustness under low-precision constraints, with a mean Quantization Stability Score (QSS) of 0.91, compared to 0.82–0.87 for existing approaches. Hardware-specific optimizations provide additional gains, such as a 2.1 × speedup on Raspberry Pi 4, reduced latency on Jetson Xavier and lower energy consumption on Intel NCS2. The results of Ablation studies affirm that dynamic quantization, hardware feedback, and stability and mechanisms are an absolute necessity, improving accuracy by up to 3.1% and decreasing latency by 31% over sequential optimization. These results prove the efficiency of dynamic quantization-aware NAS to classify encrypted traffic and emphasize its prospects in broader 5G edge AI applications.

3 July 2025
Verified
Conference Paper
Closed access
HAHaider Abd Alrazaq
Haider Abd Alrazaq
RSRavinder Sharma
Ravinder Sharma
SUSherkhanov Sultonmurod Davronboy Ugli
Sherkhanov Sultonmurod Davronboy Ugli
Abdullah Ghanim JaberAbdullah Ghanim Jaber
Abdullah Ghanim Jaber