Hafzullah Is

This study examines the levels of artificial intelligence (AI) anxiety among academic staff and students at Batman University. Data were collected from 500 academic personnel and 500 students using the Artificial Intelligence Anxiety Scale (AIAS) and analyzed through descriptive statistical methods, including t-tests and ANOVA. The study focused on demographic variables such as gender, age, academic field, and years of experience. The findings indicate that male academic staff experienced higher levels of AI anxiety, particularly in job change, sociotechnical blindness, and AI structuring. Less experienced academics also reported elevated anxiety. In the student group, females and those studying in social sciences showed higher anxiety, with the highest job change anxiety detected among the 17–25 age group. These results suggest that AI anxiety varies significantly across demographic categories. The study emphasizes the need for targeted awareness and training programs within academic institutions to support adaptation to AI technologies.

This research rigorously investigates the cybersecurity frameworks within academic institutions, emphasizing the pivotal role and security of Active Directory (AD) systems. By conducting an in-depth analysis of AD infrastructures across 12 universities with critical digital environments, this study scrutinizes access control mechanisms, user identity management, and network segmentation strategies. The findings reveal profound security lapses, such as excessive administrative privileges granted to 75% of non-administrative users and the absence of Demilitarized Zones (DMZs) in 80% of the institutions. Additionally, 65% of the institutions exhibited critical vulnerabilities by not integrating public devices, such as printers and laboratory computers, into the AD framework. The study further highlights the escalating threat of cryptolocker and ransomware attacks, which have increasingly targeted institutions, resulting in significant data encryption and operational disruptions. Moreover, challenges related to the deployment and management of advanced cybersecurity solutions, like CrowdStrike, underscore the complexities in maintaining up-to-date defenses. These issues are compounded by frequent update and upgrade failures, adversely impacting AD health and overall network security. This paper delineates strategic recommendations to enhance AD security, supported by empirical evidence showing a 92% improvement in defense against cyber attacks upon implementing these measures. The insights garnered from this study are aimed at fortifying the cybersecurity postures of academic institutions, thereby mitigating the escalating threats in the digital landscape.

This study underscores the critical importance of computer energy efficiency within institutions, using advanced tools to passively gather information for a rigorous analysis of computing devices' energy consumption. By examining a diverse array of over 2,500 devices across 40 institutions, my methodology combines statistical analyses with machine learning algorithms to identify patterns and outliers in energy use. The findings advocate for a strategic selection of energy-efficient devices and topologies, demonstrating a potential reduction in energy consumption by over 30% without sacrificing performance. I propose a holistic approach that includes adopting optimal device combinations and energy-efficient network topologies, integrating renewable energy sources, and utilizing advanced cooling solutions to further enhance sustainability. This comprehensive strategy not only aligns with global sustainability efforts but also offers a blueprint for institutions aiming to optimize their technology infrastructure for energy efficiency. Our study provides actionable insights and a framework for continuous energy consumption monitoring, emphasizing the importance of adaptive strategies in the face of evolving technological landscapes.

Amidst the growing sophistication of phishing threats that exploit human vulnerabilities, this study investigates the effectiveness of Security Awareness Training (SAT) enhanced by Large Language Models (LLMs). Targeting a diverse group of 1,270 participants, including academicians, officers, and students, it aims to evaluate whether LLM-driven SAT can strengthen phishing defenses and cultivate a more resilient digital environment. Initial assessments revealed a baseline Phish Prone Percentage (PPP) of 18.3%, indicating a pronounced vulnerability across participant groups. The deployment of an LLM-enhanced SAT program, characterized by its adaptive and interactive training modules, led to a significant post-training reduction in PPP to 6.3%. This outcome demonstrates the program's success in mitigating phishing risks and underscores the necessity of evolving SAT strategies to combat the dynamic nature of phishing attacks. The study's findings, illustrating a substantial improvement in phishing defense capabilities through LLM-integrated SAT, advocate for the integration of advanced technologies in cybersecurity education. By effectively lowering phishing vulnerability from 18.3% to 6.3%, this research highlights the critical role of innovative training methodologies in enhancing digital security across varied academic and professional landscapes.

The $5 billion update error in CrowdStrike’s security software led to global disruptions, affecting airports, hospitals, and banking systems. This issue, caused by a faulty software update, resulted in Microsoft Windows computers experiencing "blue screen" failures, impacting approximately 8.5 million devices globally and requiring manual restarts. The malfunction halted aviation, disrupted healthcare services, and disabled some TV channels. Insurance company Parametrix estimated $5.4 billion in losses for 25% of affected Fortune 500 companies in the US and around $15 billion globally. This paper examines the cybersecurity risks associated with vulnerabilities introduced by system updates, with a focus on critical infrastructures. To assess these risks, vulnerability scans were conducted across 12 critical infrastructure organizations, revealing an average 27% vulnerability rate related to updates. Through this study, we identify the evolving threat landscape and propose mitigation strategies to enhance cybersecurity posture, targeting a performance improvement of over 90%.

Deep learning can provide successful results in detecting manipulations made with bot accounts on the X (Twitter) platform and preventing disinformation. As a method for applying deep learning algorithms, numerical values in the data set can be visualized and classified by making sense of them. The methodology used in the data visualization process and the structure of the shape created are critical for deep learning and classification performance. In this article, the effect of the format used to visualize the data in the dataset to be applied deep learning on the classification performance has been demonstrated with more than 400 experimental studies. From visualization methods; The effect of deep learning applied to images created with formats such as area graph, qrcode, spectrogram graph, color map, histogram and distribution on the classification performance with Convolutional Neural Network algorithms has been demonstrated comparatively. As a result of experimental studies, it was observed that while the highest performance was achieved with 98.67% depending on the visual and algorithm used, the performance decreased to 23.70% when different visuals and algorithms were used. In this regard, it has been determined that the methodology applied in revealing the profile confidence index, the effect of the variables that make up the data set and the algorithms used are very important, but the data visualization method used in creating the figure is important. has a critical impact on performance.

This article considers methodological approaches to determine and prevent social media manipulation specific to Twitter. Behavioral analyses of Twitter users were performed by using their profile structures and interaction types, and Twitter users were classified according to their effect size values by determining their asset values. User profiles were classified into three different categories, namely popular-active, observer-passive, and spam-bot-malicious by using k-nearest neighbor (K-NN), support vector machine (SVM), and artificial neural network (ANN) algorithms. For classification, the study used the basic characteristics of users, such as density, centralization, and diameter, as well as suggested time series such as the simple moving average and cumulative moving average. The highest accuracy was obtained by the K-NN algorithm. The results obtained with K-NN for all classes were higher than the F1-Score values obtained for the other algorithms. According to the results obtained, classification accuracy values were found to reach a maximum of 96.81% and a minimum of 92.33%. Our classification results showed that the proposed method was satisfactory for popular-active, observer-passive, and spam-bot-malicious account separation.

With the development of information technologies and the increase of means and facilities for accessing the internet, the number of users of popular social networks has started to increase rapidly. Twitter is the most requested microblog site with 336 million active users. On Twitter, users can instantly express ideas, emotions and reactions with tweets. Scalable data from microblog sites, where fast and effective responses can be obtained, can be used in political, social and economic areas. It is possible to analyze the characteristics, trends and behaviors of the users by revealing their interaction. In recent years, especially the emotional analysis of users has become very popular. By analyzing the characters, effects and trends of users in Twitter, business intelligence applications can be developed and individual and social strategies can be created.In this study, the Twitter profiles of Turkey presidential election candidate were examined. Profiles was measured by physical and emotional metrics such as density, reciprocity, centralization, modularity, polarity, subjectivity. Users profiles confidence level and social authority indexes was determined. Relations with each other have been revealed. By examining these metrics, the effects of social media usage habits on users' followers are examined. Social media performances were measured from each candidate's tweets and follower analysis. Confidence indices related to 6-week time series were extracted using hierarchical classification algorithm.

During last 10 years, internet usage network has spread at an unforeseen rate. Concurrently, many services including official transactions have been granted from the internet. By the way, it has become an area where internet users have access to apps and interact with each other over social media. People's internet usage habits have become a domain that can give information about the areas of the population and their interests. In social networks, group analysis and connection predictions are popular terms that construct base of a lot of scientific works. This terms, especially, used for; ads customize, habits and tendency retain or determination of friendship based on same character. Network users can be grouped according to their activity and character, or they can be grouped according to their movement over the network by their estimation algorithms. In this study, all dataset filtered to get anonymous traffic logs that generated by a part of internet users who are in the network of an institution and the community was explored. All logs analyzed according to Category-based diversity, time periods of accesses and usage periods. As expected, communities were established from the habits of users with “Pattern Based Frequency Analysis Method”. Voluntary experimental users whose mac addresses were already defined and belong to determined groups redistributed with specified method. Lastly, via calculating the achievement of correct distribution results the success of method found out.