Karrar Ibrahim Abdulameer
Curriculum Vitae
Research Interests
Published researchs
الموقع الالكتروني
karrar.ibrahim@alzahraa.edu.iq
Doctor of Philosophy (Ph.D.) in Internetworking Technology
Networking, Network security, 5G, AI, Communications
There are no published researchs
PortScan attacks are a common security threat in computer networks, where an attacker systematically scans a range of network ports on a target system to identify potential vulnerabilities. Detecting such attacks in a timely and accurate manner is crucial to ensure network security. Attackers can determine whether a port is open by sending a detective message to it, which helps them find potential vulnerabilities. However, the best methods for spotting and identifying port scanner attacks are those that use machine learning. One of the most dangerous online threats is PortScan attack, according to experts. The research is work on detection while improving detection accuracy. Dataset containing tags from network traffic is used to train machine learning techniques for classification. The JRip algorithm is trained and tested using the CICIDS2017 dataset. As a consequence, the best performance results for JRip-based detection schemes were 99.84%, 99.80%, 99.80%, and 0.09 ms for accuracy, precision, recall, F-score, and detection overhead, respectively. Finally, the comparison with current models demonstrated our model's proficiency and advantage with increased attack discovery speed. © 2023, Institute of Advanced Engineering and Science. All rights reserved.
Distributed denial of service (DDoS) attacks have been identified as one of the greatest threats to software-defined networking (SDN) because they are highly effective, hard to detect, and easy to use, and they take advantage of vulnerabilities in which the new architecture still exists. This paper describes one technique for denying the RYU controller’s services, which can cause the controller’s resources to be depleted if a significant number of packets from various zombie hosts are sent to the controller using spoofed source internet protocol (IP) addresses. In order to demonstrate the impact of the attack, we measure various metrics related to RYU controllers such as its central processing unit (CPU) usage and network throughput. In this work, Mininet was used to simulate the data plane and measure metrics such as random access memory (RAM) usage, CPU load, and link latency. © 2023 Institute of Advanced Engineering and Science. All rights reserved.
Precise and efficient classification of histological cell nuclei is of great significance because of its promising application in the domain of medical image analysis. It facilitates the physician to explore various factors and better understand the treatment of cancer. Due to cellular heterogeneity detection and classification of cell nuclei in histopathology images of tissue stained with the typical haematoxylin and eosin stain becomes a tedious process. The deep learning approach has been demonstrated to produce remarkable outcomes on histopathology images in different fields. Therefore, this study designs an Automated Routine Colon Cancer Nuclei Classification Using Black Widow Optimization with Deep Learning (ARCCNC-BWODL) model. The presented ARCCNC-BWODL algorithm focuses majorly on the identification and classification of RCC cell nuclei. The presented algorithm applies improved Faster SqueezeNet model to make feature vectors. Besides, the hyperparameter tuning of the Faster SqueezeNet approach is performed via the BWO system. To classify the nuclei effectively, long short term memory is used. The simulation outcome of the ARCCNC-BWODL model was tested on medical imaging database and the outcomes exhibited the enhancements of the ARCCNC-BWODL over other DL techniques. © 2023 IEEE.
Nowadays, Text mining technology is broadly utilized for a wide range of business government, and research needs. Fake news recognition intends to assist users to divulge various fake news. Fake news on several vaccinations, medicines, and foods concerning the COVID19 pandemic has augmented intensely. Such fake news reports make people trust false and harmful stories and claims, and they even affect the vaccination opinions of people. Immediate detection of COVID19 fake news may assist to decrease the spread of confusion fear, and health risks among residents. Text mining related data analytics techniques are employed to detect COVID19 based fake news on social networking sites. This article introduces an African Vulture Optimization with Deep Learning based COVID-19 Fake News Detection (AVODL-FND) method. The goal of the AVODL-FND model lies in the identification of fake news relevant to COVID19 circulated in social media. In the presented AVODL-FND method, the initial stage of data preprocessing is carried out in three distinct levels to transform the input data into a useful format. The presented AVODL-FND model applies multi-head attention-based bidirectional long short-term memory (MHA-BLSTM) approach for the detection of fake news. For enhancing the detection performance of the MHA-BLSTM model, the AVO algorithm is utilized for the hyperparameter tuning process. The experimental outcomes of the AVODL-FND model on the fake news dataset show the significant performance of the AVODL-FND model over other DL methods. © 2023 IEEE.
Data mining is a process that interacts with a large dataset to determine complex interesting patterns from unknown structured data. Visual data mining (VDM) is considered a combination of two disciplines: data mining and visualization to explore useful and implicit knowledge from a huge dataset. Also, it is strongly associated with multimedia systems, high-performance computing, computer graphics, human-computer interaction, and pattern recognition. Recently, VDM approaches find useful in the healthcare sector to aid the decision-making process. In addition, the design of VDM approaches for healthcare applications needs special consideration to make sure security of data. Artificial intelligence (AI) methods play a vital role in achieving scalability and accurate analysis from real-time environments. Therefore, this study develops a blockchain-assisted quantum bacterial colonial optimization with deep learning (BAQBCO-DL) algorithm for VDM in the healthcare environment. The proposed BAQBCO-DL model exploits BC technology for secured data communication in the healthcare sector. In addition, the BAQBCO-DL model designs a U-Net-based segmentation with an optimal densely connected network (DenseNet) model for feature extraction, and the hyperparameter tuning process is performed using the QBCO algorithm. Finally, a probabilistic neural network (PNN) classifier is utilized to define appropriate class labels. An extensive experimental analysis is conducted to report the superior performance of the BAQBCO-DL technique and the outcomes are measured under several aspects. The simulation outcomes underlined the supremacy of the BAQBCO-DL algorithm over recent methodologies. © 2023 IEEE.
In recent times, vehicular ad hoc network (VANET) has been significantly considered by several service providers from urban locations. Such a network could not only prevent accidents and enhance road safety but among them offer a means of entertainment to passengers. But, based on the present analysis, effective routing is until remained a big open problem for the VANETs. The vehicle's limited broadcast range reasons the route for participating from data broadcast in the source S to destination D can vanish seldom. So, the stable route from source to destination requires for delivering data packets at planned destination. This study designs an Intelligent Artificial Humming Bird Optimization based Distance Aware Routing (IAHBO-DAR) scheme for VANET. The presented IAHBO-DAR model mainly considered the distance metric for route selection process. The IAHBO-DAR model is based on the integration of standard AHBO algorithm with dynamic oppositional based learning (DOBL) concept. In addition, the IAHBO-DAR model computes a fitness function with three variables like node degree (ND), residual energy (RE), and distance. A wide range of simulations have been conducted to exhibit the enhanced routing performance of the IAHBO-DAR approach. The experimental value demonstrates the betterment of the IAHBO-DAR algorithm compared to recent approaches. © 2023 IEEE.
Intelligent Transportation system becomes a hot research topic, afterwards Internet of Things (IoT)-based sensors have been efficiently integrated with vehicular Adhoc networks (VANET). Routing is a challenging issue in VANET to enable the effectual communication of data. Computational Intelligence (CI) is an emerging field that applies or develops a set of nature-inspired algorithms and methodologies to overcome challenging problems in real-time. The algorithms and methodologies are more commonly applied to manage complex engineering problems for which the traditional or mathematical modelling method is ineffective. Therefore, this study develops an Improved Coot Optimization Algorithm based Energy Aware Multi-Hop Routing Protocol (ICOA-EAMHRP) for VANET. The proposed ICOA-EAMHRP model is concentrated on the effectual detection of optimal routes in a short duration by the use of distinct VANET parameters. The proposed ICOA-EAMHRP model initially clusters the vehicles using a weighted-based clustering (WBC) scheme. For the route selection process, the ICOA-EAMHRP model follows the concept of two distinct modes of bird movement on the water surface such as regular and irregular. In addition, the ICOA-EAMHRP model derives a fitness function involving distinct input parameters such as Euclidean distance, mobility, bandwidth, and energy. To validate the improved performance of the ICOA-EAMHRP model, a detailed experimental analysis is carried out. The comparative study reported the enhanced performance of the ICOA-EAMHRP model over existing routing protocols. © 2023 IEEE.
With the recent advancements in 3D acquisition technology, 3D sensors becomes ubiquitous and cheap, comprising different kinds of 3D laser scanners (or LiDAR) on diverse platforms. The application of deep learning (DL) models on 2D images gained significant interest in the area of computer vision applications. Recently, the DL models for point clouds (PC) begins to draw considerable attention. But it is still needed to deeply understand the role of DL and geometry domain knowledge to understand PC. Therefore, this study develops an Automated Wood and Leaf Classification using Coyote Optimization Algorithm with Deep Learning (AWLC-COADL) Model for Terrestrial LiDAR PC. The presented AWLC-COADL technique mainly aims to classify the leaf and woody elements on the TLS data. The precise distinction between leaf and woody elements is important to estimate the leaf area index (LAI) and wood area index (WAI). To accomplish this, the AWLC-COADL technique applies the PointNet model for object detection in 3D PC. In addition, the wood and leaf classification process can be performed by the use of bidirectional long short term memory. At last, the COA is exploited to adjust the hyperparameter values of the BiLSTM method to enhance its efficiency. A widespread simulation analysis is performed to assure the significance of the AWLC-COADL model. A brief comparison results reported the promising results of the AWLC-COADL model over other existing models. © 2023 IEEE.
Pancreatic cancer is a deadly form of tumor and estimations are reduced in the present scenarios. Automatic pancreatic cancer classification by the use of computer-aided diagnosis (CAD) technique becomes essential for tracking, predicting, and classifying the presence of pancreatic cancer. Artificial intelligence (AI) techniques assist in medical decision-making with the consideration of huge amounts of medical imaging data. The latest improvements in deep learning (DL) algorithms assist in the effective design of pancreatic cancer classification models. In this view, this paper establishes an end-to-end pancreatic cancer classification using a metaheuristic and deep transfer learning (ETEPCC-MDTL) algorithm. The proposed ETEPCC-MDTL algorithm aims to precisely determine the presence of pancreatic tumors on computed tomography (CT) images. To accomplish this, the ETEPCC-MDTL method initially applies bilateral filtering (BF) based pre-processing and Shannon's entropy-based image segmentation technique. In addition, Neural Architectural Search Net (NASNet) model is employed for feature extraction purposes with cat swarm optimizer (CSO) as a hyperparameter optimization algorithm. Moreover, glowworm swarm optimization (GSO) with Elman neural network (ENN) is exploited for the classification process. The experimental validation of the ETEPCC-MDTL algorithm is tested under benchmark medical image databases. © 2023 IEEE.