Suhad Faisal Behadili is a professor in the Department of Computer Science in the College of Science at the University of Baghdad, Baghdad/Iraq.
She has a Ph.D. from the LITIS at Normandie University - Le Havre/ France. She currently editorial member of the American Journal of Information Science and Technology. She is a program committee member in several international conferences, and a reviewer for IJS and IASET journals. She has published numerous technical papers, undergraduate/Postgraduate teaching, and outreach.
EDUCATION
PhD in Communication and Networking (LITIS)
Université du Havre France: Le Havre, Haute-Normandie, FR
RESEARCH INTERESTS
Data Mining
Mobile communication and wireless networks
mobility Networking and Communications
Remote Sensing
urban studies
29
Scopus Publications
Scopus Publications
One-Class and Multi-Class Malware Classification Using Hybrid and Supervised Machine Learning Techniques Mohammed Saadoon, Suhad Faisal Behadili Iraqi Journal of Science, 2026 Cybercriminals or hackers design malware programs with malicious intent to steal, spy, and destroy victim's computers. Malware encompasses various forms, such as viruses, trojans, ransomware, spyware, and adware, each requiring effective classification for accurate identification and mitigation. High-quality datasets are crucial for training classification models, and the CIC-MalMem-2022 dataset, containing 58,596 records and 55 numerical features, is an essential resource in this regard. This study used artificial intelligence techniques and algorithms such as KNN, Decision Tree, Random Forest, SVM, and Naïve Bays, as well as supervised and hybrid machine learning by integrating Random Forest and K-Nearest Neighbors KNN to improve classification performance with one-class classification and multi-class. In terms of accuracy score, the best results achieved by the proposed methods were random forest with 99.98%, hybrid random forest + K-Nearest Neighbors with 99.93%, decision tree with 99.95%, k-Nearest Neighbors with 99.87%, support vector machine with 99.80%, and naïve bays with 98.90%. These metrics (accuracy, precision, and recall) reflect the models' effectiveness in classifying instances. Accuracy measures overall correctness, precision evaluates the quality of positive predictions, and recall assesses the ability to identify true positives. The consistently high scores demonstrate the reliability and robustness of these methods for malware classification.
Fingerprint Forgery Detection and Person Identification Based on Deep Learning Mohammed Abdul Ameer Jabbar, Abdulkareem Merhej Radhi, Sabreen A.Zahra Mghames, Suhad Faisal Behadili, Muneera Alsaedi Iraqi Journal of Science, 2025 Automated fingerprint recognition and authentication have been widely used in biometrics applications and as a personal identity tool due to their dependability and unique characteristics. The person's fingerprint must be authentic and not altered or forged to be used to verify that person's identity. It is more difficult to determine whether a fingerprint is real /authenticated. The presented work aims to design two models based on a convolutional neural network (CNN) with the ability to detect whether the fingerprints are authenticated or not. The proposed methodology includes two levels: the first involves forgery detection of fingerprints. Whereas the second level exam ines fingerprint identification. Furthermore, a reliable deep learning technique that includes Transfer Learning (TL) and building the architecture of CNN from scratch was utilized to diagnose and identify fingerprints for 100 persons using the SOCO dataset. Thus, the results recorded higher accuracy at 98.69% and 99.08% sensitivity of forgery detection. Furthermore, it achieved an optimal rate for matching fingerprints and outperformed other TL models (VGG16, VGG19, ResNet50) and related works. For this reason, it could be considered a successful model for Biometric fingerprint authentication and forgery detection.
Mobility Prediction Based on LSTM Multi-Layer Using GPS Phone Data Nabaa Mhalhal, Suhad Behadili Iraqi Journal for Electrical and Electronic Engineering, 2025 Precise Prediction of activity location is an essential element in numerous mobility applications and is especially necessary for the development of tailored sustainable transportation systems. Next-location prediction, which involves predicting a user’s future position based on their past movement patterns, has significant implications in various domains, including urban planning, geo-marketing, disease transmission, Performance wireless network, Recommender Systems, and many other areas. In recent years, various predictors have been suggested to tackle this issue, including state-of-the-art ones that utilize deep learning techniques. This study introduces a robust Model for predicting the future location path of a user based on their known previous locations. The study proposes the use of a Long Short-Term Memory (LSTM) prediction scheme, which is well-suited for learning from sequential data; then a fully connected neuron is employed to decrease the sparsity of the data, resulting in accurate predictions for the path of the user’s next location. The suggested strategy demonstrates superior prediction accuracy compared to a state-of-the-art method, with improvements of up to a loss error of 0.002 based on real-life datasets (Geolife). The results demonstrate that the reliability of forecasts is excellent, indicating the accuracy of the predictions.
A Review of Algorithms and Platforms for Offloading Decisions in Mobile Cloud Computing Fatima Haitham Murtadha, Suhad Faisal Behadili Iraqi Journal for Electrical and Electronic Engineering, 2025 With the substantial growth of mobile applications and the emergence of cloud computing concepts, therefore mobile Cloud Computing (MCC) has been introduced as a potential mobile service technology. Mobile has limited resources, battery life, network bandwidth, storage, and processor, avoid mobile limitations by sending heavy computation to the cloud to get better performance in a short time, the operation of sending data, and get the result of computation call offloading. In this paper, a survey about offloading types is discussed that takes care of many issues such as offloading algorithms, platforms, metrics (that are used with this algorithm and its equations), mobile cloud architecture, and the advantages of using the mobile cloud. The trade-off between local execution of tasks on end-devices and remote execution on the cloud server for minimizing delay time and energy saving. In the form of a multi-objective optimization problem with a focus on reducing overall system power consumption and task execution latency, meta-heuristic algorithms are required to solve this problem which is considered as NP-hardness when the number of tasks is high. To get minimum cost (time and energy) apply partial offloading on specific jobs containing a number of tasks represented in sequences of zeros and ones for example (100111010), when each bit represents a task. The zeros mean the task will be executed in the cloud and the ones mean the task will be executed locally. The decision of processing tasks locally or remotely is important to balance resource utilization. The calculation of task completion time and energy consumption for each task determines which task from the whole job will be executed remotely (been offloaded) and which task will be executed locally. Calculate the total cost (time and energy) for the whole job and determine the minimum total cost. An optimization method based on metaheuristic methods is required to find the best solution. The genetic algorithm is suggested as a metaheuristic Algorithm for future work.
