Papers by vaman Atroushi
Academic Journal of Nawroz University
Sending digital pictures across open networks has emerged as a major privacy risk in recent years... more Sending digital pictures across open networks has emerged as a major privacy risk in recent years. Despite the environment's adaptability and the many benefits it offers, there are, however, a significant number of threats to one's privacy and safety. A great number of cryptosystems have been proposed in the literature on picture encryption in an effort to make communication more secure. For the purpose of data transmission, the AES algorithm is utilized due to the increased efficiency it offers in the block. This paper proposes image encryption techniques based on the AES algorithm and Henon map. The plain image has been encrypted using the AES algorithm at the first step. Then, the Henon map is used to generate a random key which is required to provide the second step of encryption. This step of encryption has been performed using the XOR operation. The results of the studies demonstrate that the strategy proposed resolves the issues that are present in conventional techn...
Credit Card Fraud Detection using KNN, Random Forest and Logistic Regression Algorithms : A Comparative Analysis
Indonesian journal of computer science/Indonesian Journal of Computer Science, Feb 6, 2024
Academic Journal of Nawroz University (AJNU),, 2023
Sending digital pictures across open networks has emerged as a major privacy risk in recent years... more Sending digital pictures across open networks has emerged as a major privacy risk in recent years. Despite the environment's adaptability and the many benefits it offers, there are, however, a significant number of threats to one's privacy and safety. A great number of cryptosystems have been proposed in the literature on picture encryption in an effort to make communication more secure. For the purpose of data transmission, the AES algorithm is utilized due to the increased efficiency it offers in the block. This paper proposes image encryption techniques based on the AES algorithm and Henon map. The plain image has been encrypted using the AES algorithm at the first step. Then, the Henon map is used to generate a random key which is required to provide the second step of encryption. This step of encryption has been performed using the XOR operation. The results of the studies demonstrate that the strategy proposed resolves the issues that are present in conventional techniques of encryption. The histogram of the encryption picture is consistently spaced despite being different from the histogram of the original image. The recommended approach is extremely sensitive to the key value; even minute adjustments result in a distinct visual representation. As a result, apps that provide real-time picture encryption while operating over unsecured networks are suitable for the unique technique.
ICONTECH INTERNATIONAL JOURNAL, 2021
In recent years, it has become increasingly common for individuals to connect with their relative... more In recent years, it has become increasingly common for individuals to connect with their relatives and friends, read the most recent news, and discuss various social topics using online social platforms such as Twitter and Facebook. As a consequence of this, anything that is considered spam can quickly spread among them. Spam identification is widely regarded as one of the most significant problems in text analysis. Previous studies on detecting spam concentrated primarily on English-language content and paid little attention to other languages. The information gathered by the University of California; Irvine served as the basis for developing our spam detection technology (UCI). This study investigates the effectiveness of various supervised machine learning algorithms, such as the J48, K-Nearest Neighbors (KNN), and Decision Tree (DT), in identifying spam and ham communications. SMS spam is becoming more widespread as the number of internet users continues to rise, and many businesses disclose their customers' personal information. E-mail spam filtering is the progenitor of SMS spam filtering, which inherits many features. We evaluate the proposed method based on accuracy, recall, and precision. Experiments showed that the Decision Tree method obtained higher accuracy than other machine learning classifiers.
applied computing journa, 2022
Author(s) and ACAA permit unrestricted use, distribution, and reproduction in any medium, provide... more Author(s) and ACAA permit unrestricted use, distribution, and reproduction in any medium, provided the original work with proper citation. This work is licensed under Creative Commons Attribution International License (CC BY 4.0).
International Journal of Mathematics, Statistics, and Computer Science (IJMSCS), 2023
Recognition of facial expressions has been an important topic of study over the last several deca... more Recognition of facial expressions has been an important topic of study over the last several decades, and despite the advancements that have been made, it is still difficult to do because of the significant intra-class diversity. The handcrafted feature is used in traditional methods to address this issue. This feature is then preceded by a classifier that is trained using a database of pictures or videos. Most of these works do quite well on datasets of photographs that were recorded in a controlled environment. However, they do not perform as well on datasets that are more difficult to work with since they include greater image variance and partial faces. The Histogram of Oriented Gradient (HOG) descriptor is the foundation for the strategy that is suggested in this study. During the initial step of the procedure, the input picture is pre-processed to locate the datum region, which assists in the extraction of the most relevant characteristics. After that, the Random Forest (RF) algorithm was employed as a classifier for facial expressions. The Japanese Female Facial Emotions Database (JAFFE) is used to assess our technique. The experimental findings demonstrated that the suggested method is accurate and effective in identifying facial expressions.
Because credit cards are utilized so frequently, fraud appears to be a significant concern in the... more Because credit cards are utilized so frequently, fraud appears to be a significant concern in the credit card industry. It is challenging to quantify the effects of misrepresentation. Globally, credit card fraud has cost institutions and consumers billions of dollars. Despite the existence of numerous anti-fraud mechanisms, fraudsters continue to seek out novel methods and strategies to commit fraud. An additional challenge in the estimation of credit card fraud loss is that the magnitude of unreported or undetected forgeries cannot be determined, only losses associated with those frauds that have been detected can be measured. Implementing effective fraud detection algorithms through the utilization of machine-learning techniques is crucial in order to mitigate these losses and provide support to fraud investigators. This paper presents a machine learning-based method for the detection of credit card fraud. Three methodologies are implemented on the raw and pre-processed data. Python is used to implement the work. By comparing the accuracy-based performance evaluations of k-nearest neighbor and logistic regression with Random Forest.
Conference Presentations by vaman Atroushi
Springer, 2024
Preprocessing medical datasets adequately is crucial for improving the
accuracy and dependability... more Preprocessing medical datasets adequately is crucial for improving the
accuracy and dependability of machine learning models, especially for healthcare applications. This study uses the popular Diabetes dataset to compare several preprocessing techniques for K-Nearest Neighbors (KNN) classification. Three preprocessing methods—Min-Max Scaling, Standardization, and Robust Scaling— were selected. These techniques are vital for reducing the difficulties of dealing with inconsistent, incomplete, and noisy medical data, which affects the accuracy of classification models’ predictions and diagnoses. Each preprocessing approach is thoroughly examined in the research, with special emphasis on the normalizing capabilities of Min-Max Scaling, the resilience of Robust Scaling to outliers, and the standardization accomplished by Standardization. This study uses the KNN algorithm to forecast the occurrence of diabetes mellitus, with accuracy being the
primary indicator of performance. Findings from the comparative analysis shed light on the subtle effects of each preprocessing method on the KNN classifier’s prediction accuracy, offering helpful information for improving the classification model for diabetes diagnosis. The work aims to further the development of trustworthy and understandable models for healthcare applications by highlighting the need of careful preprocessing of medical datasets.
Uploads
Papers by vaman Atroushi
Conference Presentations by vaman Atroushi
accuracy and dependability of machine learning models, especially for healthcare applications. This study uses the popular Diabetes dataset to compare several preprocessing techniques for K-Nearest Neighbors (KNN) classification. Three preprocessing methods—Min-Max Scaling, Standardization, and Robust Scaling— were selected. These techniques are vital for reducing the difficulties of dealing with inconsistent, incomplete, and noisy medical data, which affects the accuracy of classification models’ predictions and diagnoses. Each preprocessing approach is thoroughly examined in the research, with special emphasis on the normalizing capabilities of Min-Max Scaling, the resilience of Robust Scaling to outliers, and the standardization accomplished by Standardization. This study uses the KNN algorithm to forecast the occurrence of diabetes mellitus, with accuracy being the
primary indicator of performance. Findings from the comparative analysis shed light on the subtle effects of each preprocessing method on the KNN classifier’s prediction accuracy, offering helpful information for improving the classification model for diabetes diagnosis. The work aims to further the development of trustworthy and understandable models for healthcare applications by highlighting the need of careful preprocessing of medical datasets.
accuracy and dependability of machine learning models, especially for healthcare applications. This study uses the popular Diabetes dataset to compare several preprocessing techniques for K-Nearest Neighbors (KNN) classification. Three preprocessing methods—Min-Max Scaling, Standardization, and Robust Scaling— were selected. These techniques are vital for reducing the difficulties of dealing with inconsistent, incomplete, and noisy medical data, which affects the accuracy of classification models’ predictions and diagnoses. Each preprocessing approach is thoroughly examined in the research, with special emphasis on the normalizing capabilities of Min-Max Scaling, the resilience of Robust Scaling to outliers, and the standardization accomplished by Standardization. This study uses the KNN algorithm to forecast the occurrence of diabetes mellitus, with accuracy being the
primary indicator of performance. Findings from the comparative analysis shed light on the subtle effects of each preprocessing method on the KNN classifier’s prediction accuracy, offering helpful information for improving the classification model for diabetes diagnosis. The work aims to further the development of trustworthy and understandable models for healthcare applications by highlighting the need of careful preprocessing of medical datasets.