Mahmoud Baklizi
Current Organizations / Institutes
The organizations/institutes from the scholar's latest article
Research Subjects
Mahmoud Baklizi has a total of 19 co-authors. Here are the top co-authors by number of shared publications:
- Mohammad Alkhazaleh
Isra University, Amman, Jordan
4 shared publications - Jamal Zraqou
University of Petra, Jordan
3 shared publications - Faisal Alzyoud
Isra University
2 shared publications - Issa Atoum
Department of Software Engineering, Philadelphia University, Amman 19392, Jordan
2 shared publications - Musab Bassam Al-Zghoul
Isra University, Amman, Jordan
2 shared publications - Adi Maaita1 shared publication
Latest Publications
- 1 December 2026
- Published by Institute of Advanced Engineering and Science in Bulletin of Electrical Engineering and Informatics
- Vol. 15 (6) , 5514-5531
A leakage-audited comparison of neural architectures for fake news detection: a cross-dataset evaluation
Reported accuracies for fake news detection span roughly 60% to 99%, too wide a spread to be explained by architecture. We trace much of it to leakage. Auditing a corpus built from word embedding over linguistic features for fake (WELFake), LIAR, FakeNewsNet, and a published COVID-19 dataset, we find five publisher artifacts, chiefly wire-service datelines and image-credit lines, in 31,701 of the 96,626 documents collected (32.8%); they predict the label at about 99.7% purity by string matching alone, and fake articles are exact duplicates fifteen times more often than real ones. They survive preprocessing and are not removed by source-aware splitting, because they sit inside the article. After mitigating both channels and partitioning 87,408 deduplicated articles by source, we compare seven architectures under three seeds each and detect no robust separation: the spread between their macro-averaged accuracies (2.16 points) is smaller than the 2.63 expected from seed variation alone, an absence of detected separation rather than evidence of equivalence. A term frequency-inverse document frequency (TF-IDF) logistic regression (LR) fitted in 77 seconds reaches 88.05% against 89.30% for the best neural model. WELFake supplies 69.6 of those 89.3 points, and on LIAR every run gives an area under the curve (AUC) of 0.57-0.59. Predictions, splits and an audit tool are released.
- 1 December 2026
- Published by Institute of Advanced Engineering and Science in Bulletin of Electrical Engineering and Informatics
- Vol. 15 (6) , 5514-5531
A leakage-audited comparison of neural architectures for fake news detection: a cross-dataset evaluation
Reported accuracies for fake news detection span roughly 60% to 99%, too wide a spread to be explained by architecture. We trace much of it to leakage. Auditing a corpus built from word embedding over linguistic features for fake (WELFake), LIAR, FakeNewsNet, and a published COVID-19 dataset, we find five publisher artifacts, chiefly wire-service datelines and image-credit lines, in 31,701 of the 96,626 documents collected (32.8%); they predict the label at about 99.7% purity by string matching alone, and fake articles are exact duplicates fifteen times more often than real ones. They survive preprocessing and are not removed by source-aware splitting, because they sit inside the article. After mitigating both channels and partitioning 87,408 deduplicated articles by source, we compare seven architectures under three seeds each and detect no robust separation: the spread between their macro-averaged accuracies (2.16 points) is smaller than the 2.63 expected from seed variation alone, an absence of detected separation rather than evidence of equivalence. A term frequency-inverse document frequency (TF-IDF) logistic regression (LR) fitted in 77 seconds reaches 88.05% against 89.30% for the best neural model. WELFake supplies 69.6 of those 89.3 points, and on LIAR every run gives an area under the curve (AUC) of 0.57-0.59. Predictions, splits and an audit tool are released.
- Open Access
- 1 August 2026
- Published by Institute of Advanced Engineering and Science in Bulletin of Electrical Engineering and Informatics
- Vol. 15 (4) , 3483-3494
Scoring smarter: deep learning-based basketball scoring detection in real time
Basketball scoring detection is challenged by scene complexity, varying camera angles, and occlusions. This paper presents a real-time basketball scoring system that combines you only look once version 8 (YOLOv8) for hoop detection, a vision transformer (ViT) for spatiotemporal motion modeling, and deep simple online and realtime tracking (DeepSORT) for object tracking, in order to overcome the accuracy loss that classical systems suffer during dynamic gameplay. The system was evaluated on a custom dataset of 4,000 annotated images and five full-length broadcast videos containing 44 verified scoring events. The proposed YOLOv8+ViT+DeepSORT model achieved 96.84% average precision at 50% intersection over union (AP50), 95.2% precision, 93.8% recall, 94.5% F1-score, and 94.55% accuracy, while sustaining 85 frames per second (FPS). Comparison against YOLOv3 with frame differencing (baseline), you only look once version 8 nano (YOLOv8n) with bidirectional feature pyramid network (BiFPN) and global attention module (GAM) attention mechanisms, and BiFPN, GAM, and SimC2f-YOLO (BGS-YOLO) confirms the best overall balance across all indicators. ViT captures the temporal dynamics of motion while DeepSORT preserves object identity across frames. The framework is therefore well suited to sports analytics, automated highlight generation, and referee assistance systems where accuracy and real-time performance are essential.
- 1 August 2026
- Published by Institute of Advanced Engineering and Science in Bulletin of Electrical Engineering and Informatics
- Vol. 15 (4) , 3483-3494
Open AccessScoring smarter: deep learning-based basketball scoring detection in real time
Basketball scoring detection is challenged by scene complexity, varying camera angles, and occlusions. This paper presents a real-time basketball scoring system that combines you only look once version 8 (YOLOv8) for hoop detection, a vision transformer (ViT) for spatiotemporal motion modeling, and deep simple online and realtime tracking (DeepSORT) for object tracking, in order to overcome the accuracy loss that classical systems suffer during dynamic gameplay. The system was evaluated on a custom dataset of 4,000 annotated images and five full-length broadcast videos containing 44 verified scoring events. The proposed YOLOv8+ViT+DeepSORT model achieved 96.84% average precision at 50% intersection over union (AP50), 95.2% precision, 93.8% recall, 94.5% F1-score, and 94.55% accuracy, while sustaining 85 frames per second (FPS). Comparison against YOLOv3 with frame differencing (baseline), you only look once version 8 nano (YOLOv8n) with bidirectional feature pyramid network (BiFPN) and global attention module (GAM) attention mechanisms, and BiFPN, GAM, and SimC2f-YOLO (BGS-YOLO) confirms the best overall balance across all indicators. ViT captures the temporal dynamics of motion while DeepSORT preserves object identity across frames. The framework is therefore well suited to sports analytics, automated highlight generation, and referee assistance systems where accuracy and real-time performance are essential.
- Open Access
- 1 December 2025
- Published by Institute of Advanced Engineering and Science in Bulletin of Electrical Engineering and Informatics
- Vol. 14 (6) , 4626-4640
A powerful machine learning method for detecting phishing threats
Phishing threats exploit social engineering and deceptive web infrastructure to steal sensitive personal information, often by mimicking legitimate websites. With the proliferation of online services and the increasing prevalence of cybercrime, detecting phishing websites has become a critical challenge. This study presents a comprehensive machine learning (ML)-based approach for detecting phishing websites. A total of 48 discriminative features were extracted from 10,000 web pages—comprising 5,000 phishing and 5,000 legitimate sites. Nine ML classifiers were initially evaluated, including random forest (RF), support vector machine (SVM), and XGBoost. Ensemble models based on soft voting and stacking were then constructed to improve detection performance. Among the models, the soft voting classifier (VC) achieved the best performance with an accuracy and F1-score of 98.82%. The results indicate that ensemble learning offers a robust solution for the automated detection of phishing websites.
- 1 December 2025
- Published by Institute of Advanced Engineering and Science in Bulletin of Electrical Engineering and Informatics
- Vol. 14 (6) , 4626-4640
Open AccessA powerful machine learning method for detecting phishing threats
Phishing threats exploit social engineering and deceptive web infrastructure to steal sensitive personal information, often by mimicking legitimate websites. With the proliferation of online services and the increasing prevalence of cybercrime, detecting phishing websites has become a critical challenge. This study presents a comprehensive machine learning (ML)-based approach for detecting phishing websites. A total of 48 discriminative features were extracted from 10,000 web pages—comprising 5,000 phishing and 5,000 legitimate sites. Nine ML classifiers were initially evaluated, including random forest (RF), support vector machine (SVM), and XGBoost. Ensemble models based on soft voting and stacking were then constructed to improve detection performance. Among the models, the soft voting classifier (VC) achieved the best performance with an accuracy and F1-score of 98.82%. The results indicate that ensemble learning offers a robust solution for the automated detection of phishing websites.
- Open Access
- 1 December 2025
- Published by Institute of Advanced Engineering and Science in Bulletin of Electrical Engineering and Informatics
- Vol. 14 (6) , 4823-4833
Evaluating the effectiveness of Havij for structured query language injection exploitation in web applications
Structured query language injection (SQLi) is still one of the most critical risks to web application security, as it allows attackers to interfere with sensitive data and even a complete database infrastructure. Although many automated tools are available, previous studies usually achieve only descriptive briefs, which do not offer empirical assessments that measure the performance and the usability. This research fills this void by a systematic five-stage experimental analysis of the Havij automated SQLi tool under a controlled and ethical test setup. Confirmation of vulnerability, automated exploitation, data extraction and benchmarking of performance were performed as the methodology, and the results were compared against the industry standard SQLmap tool. It was found that in less than a minute Havij was able to locate the target database, scan its structure, and steal authentication credentials, which is quite efficient and user-friendly. In contrast to the literature, our work presents not only quantitative measures (time-to-exploit, request volume, and success rate) but also a qualitative evaluation (user accessibility and limitations), which gives a comprehensive evaluation. The results highlight trade-offs between the depth and accessibility, the continued dangers of SQLi in practice, and provide recommendations that developers and security experts can implement.
- 1 December 2025
- Published by Institute of Advanced Engineering and Science in Bulletin of Electrical Engineering and Informatics
- Vol. 14 (6) , 4823-4833
Open AccessEvaluating the effectiveness of Havij for structured query language injection exploitation in web applications
Structured query language injection (SQLi) is still one of the most critical risks to web application security, as it allows attackers to interfere with sensitive data and even a complete database infrastructure. Although many automated tools are available, previous studies usually achieve only descriptive briefs, which do not offer empirical assessments that measure the performance and the usability. This research fills this void by a systematic five-stage experimental analysis of the Havij automated SQLi tool under a controlled and ethical test setup. Confirmation of vulnerability, automated exploitation, data extraction and benchmarking of performance were performed as the methodology, and the results were compared against the industry standard SQLmap tool. It was found that in less than a minute Havij was able to locate the target database, scan its structure, and steal authentication credentials, which is quite efficient and user-friendly. In contrast to the literature, our work presents not only quantitative measures (time-to-exploit, request volume, and success rate) but also a qualitative evaluation (user accessibility and limitations), which gives a comprehensive evaluation. The results highlight trade-offs between the depth and accessibility, the continued dangers of SQLi in practice, and provide recommendations that developers and security experts can implement.
- Open Access
- 1 August 2025
- Published by Institute of Advanced Engineering and Science in International Journal of Electrical and Computer Engineering (IJECE)
- Vol. 15 (4) , 3913-3927
Enhancing mobile agent protection using a hybrid security framework combining pretty good protocol and code obfuscation
The security of mobile agents, which are autonomous software entities capable of migrating between computers to execute tasks, remains a critical concern in modern information technology. Cybersecurity has been a central component of this technological revolution and continues to be one of the most essential requirements for any software or platform. Despite advances in security measures, protecting mobile agents, particularly those carrying sensitive data, while they transmit over networks remains challenging. This research proposes a novel hybrid security technique, abbreviated as pretty good privacy and code obfuscation framework (PGF), which combines pretty good privacy (PGP) with code obfuscation. PGF is designed specifically to protect mobile agents, focusing on systems like Aglets. The technique aims to safeguard the integrity and confidentiality of the agent's data during transmission. Based on the mobile agent Aglets and the PGF technique, the proposed model enhances security by introducing additional protection layers during agent creation and transmission using PGP and code obfuscation. The comparative analysis demonstrated that PGF outperformed other algorithms in terms of time efficiency and security, effectively handling large data sizes through its hybrid cryptographic approach, which combines asymmetric and symmetric encryption. The model was implemented using the Aglets framework in Java development kit (JDK) and NetBeans and showed high reliability and practicality. However, its current design is tailored to Aglets, and future work could focus on adapting the model to other platforms and optimizing its resource efficiency for constrained environments.
- 1 August 2025
- Published by Institute of Advanced Engineering and Science in International Journal of Electrical and Computer Engineering (IJECE)
- Vol. 15 (4) , 3913-3927
Open AccessEnhancing mobile agent protection using a hybrid security framework combining pretty good protocol and code obfuscation
The security of mobile agents, which are autonomous software entities capable of migrating between computers to execute tasks, remains a critical concern in modern information technology. Cybersecurity has been a central component of this technological revolution and continues to be one of the most essential requirements for any software or platform. Despite advances in security measures, protecting mobile agents, particularly those carrying sensitive data, while they transmit over networks remains challenging. This research proposes a novel hybrid security technique, abbreviated as pretty good privacy and code obfuscation framework (PGF), which combines pretty good privacy (PGP) with code obfuscation. PGF is designed specifically to protect mobile agents, focusing on systems like Aglets. The technique aims to safeguard the integrity and confidentiality of the agent's data during transmission. Based on the mobile agent Aglets and the PGF technique, the proposed model enhances security by introducing additional protection layers during agent creation and transmission using PGP and code obfuscation. The comparative analysis demonstrated that PGF outperformed other algorithms in terms of time efficiency and security, effectively handling large data sizes through its hybrid cryptographic approach, which combines asymmetric and symmetric encryption. The model was implemented using the Aglets framework in Java development kit (JDK) and NetBeans and showed high reliability and practicality. However, its current design is tailored to Aglets, and future work could focus on adapting the model to other platforms and optimizing its resource efficiency for constrained environments.
- Open Access
- 27 February 2024
- Published by International Association of Online Engineering (IAOE) in International Journal of Online and Biomedical Engineering (iJOE)
- Vol. 20 (03) , 24-38
Web Attack Intrusion Detection System Using Machine Learning Techniques
Web attacks often target web applications because they can be accessed over a network and often have vulnerabilities. The success of an intrusion detection system (IDS) in detecting web attacks depends on an effective traffic classification system. Several previous studies have utilized machine learning classification methods to create an efficient IDS with various datasets for different types of attacks. This paper utilizes the Canadian Institute for Cyber Security’s (CIC-IDS2017) IDS dataset to assess web attacks. Importantly, the dataset contains 80 attributes of recent assaults, as reported in the 2016 McAfee report. Three machine learning algorithms have been evaluated in this research, namely random forests (RF), k-nearest neighbor (KNN), and naive bayes (NB). The primary goal of this research is to propose an effective machine learning algorithm for the IDS web attacks model. The evaluation compares the performance of three algorithms (RF, KNN, and NB) based on their accuracy and precision in detecting anomalous traffic. The results indicate that the RF outperformed the NB and KNN in terms of average accuracy achieved during the training phase. During the testing phase, the KNN algorithm outperformed others, achieving an average accuracy of 99.4916%. However, RF and KNN achieved 100% average precision and recall rates compared to other algorithms. Finally, the RF and KNN algorithms have been identified as the most effective for detecting IDS web attacks.
- 27 February 2024
- Published by International Association of Online Engineering (IAOE) in International Journal of Online and Biomedical Engineering (iJOE)
- Vol. 20 (03) , 24-38
Open AccessWeb Attack Intrusion Detection System Using Machine Learning Techniques
Web attacks often target web applications because they can be accessed over a network and often have vulnerabilities. The success of an intrusion detection system (IDS) in detecting web attacks depends on an effective traffic classification system. Several previous studies have utilized machine learning classification methods to create an efficient IDS with various datasets for different types of attacks. This paper utilizes the Canadian Institute for Cyber Security’s (CIC-IDS2017) IDS dataset to assess web attacks. Importantly, the dataset contains 80 attributes of recent assaults, as reported in the 2016 McAfee report. Three machine learning algorithms have been evaluated in this research, namely random forests (RF), k-nearest neighbor (KNN), and naive bayes (NB). The primary goal of this research is to propose an effective machine learning algorithm for the IDS web attacks model. The evaluation compares the performance of three algorithms (RF, KNN, and NB) based on their accuracy and precision in detecting anomalous traffic. The results indicate that the RF outperformed the NB and KNN in terms of average accuracy achieved during the training phase. During the testing phase, the KNN algorithm outperformed others, achieving an average accuracy of 99.4916%. However, RF and KNN achieved 100% average precision and recall rates compared to other algorithms. Finally, the RF and KNN algorithms have been identified as the most effective for detecting IDS web attacks.