Pukyong National University

Phishing attacks continue to grow in scale and sophistication, causing substantial financial losses and privacy breaches worldwide. Recent advances in large language models (LLMs) have brought significant changes to the generation and detection of phishing content. This study systematically investigates the dual role of LLMs in facilitating phishing attacks and strengthening countermeasures. Using the PRISMA methodology, authors screened 142 records published between January 2023 and April 2025 and identified 36 eligible studies from major academic databases, including IEEE Xplore, ScienceDirect, ACM Digital Library, Web of Science, and Scopus. A comprehensive and rigorous analysis was conducted of research trends/themes over time, dataset characteristics, and the LLM architectures/models employed. The findings reveal that most studies relied on manually generated datasets rather than publicly available benchmark datasets, and that GPT-based models received considerably more attention than other LLM architectures. The review demonstrates that LLMs substantially enhance the generation of phishing content by producing coherent, contextually relevant, and persuasive email and website content. This capability lowers the technical barrier for attackers and potentially increases attack effectiveness. Conversely, LLMs also strengthen defensive strategies by enabling more effective analysis of textual and visual content for phishing detection. In many cases, LLM-based approaches outperform traditional machine learning and deep learning methods and, in certain contexts, approach or match human-level performance. Overall, the findings suggest that LLMs have accelerated and automated phishing-related processes, simultaneously intensifying the threat landscape and advancing defensive capabilities. • The first in-depth study to review LLMs usage in Phishing attack generation and detection. • The study reveals LLMs have accelerated and automated phishing-related processes, elevating both threats and defence mechanisms. • GenAI-based multimodal phishing attacks are on the rise due to the wider adoption of GenAI tools in general.

In this study, the use of simple attachments to existing guardrail members was investigated to evaluate their potential for reducing the overall drag force on long-span bridge decks. Wind tunnel testing and computational fluid dynamics (CFD) analysis were conducted to reveal the drag force reduction effect with the placement of attachments onto the rear sides of guardrails. A series of case studies of single-rail members and three-layer guardrails was performed to determine the optimal shape. A steel twin-box suspension bridge with a main span of 1545 m was selected as a prototype bridge deck, and the evaluation indicated an overall drag reduction ratio of 11.1% when an optimal attachment was applied. CFD analysis was also conducted to clarify the mechanisms responsible for the drag reduction achieved by the modified guardrails and to address Reynolds number effects. This study demonstrates the importance of optimizing the aerodynamic shape of guardrails and achieving drag force reduction by adding attachments to the rear sides of the guardrails without compromising the crash performance.

Background Although subjects with forward head/rounded shoulders posture (FHRSP) may exhibit changes in scapular position and periscapular muscle strength relative to normal subjects, few studies have investigated this. Objective We compared scapular position and periscapular muscle strength between subjects with and without FHRSP. Methods In total, 50 subjects with FHRSP and 50 normal subjects were included. The craniovertebral angle (CVA) and rounded shoulder angle (RSA) were measured to determine FHRSP. To compare scapular position, scapular protraction, elevation, upward rotation, and anterior tilt were measured. To compare periscapular muscle strength, the strengths of the upper trapezius, middle trapezius, lower trapezius, and serratus anterior were measured using an isometric hand-held dynamometer. Results Relative to normal subjects, those with FHRSP had elevated, protracted, and anteriorly tilted scapulae and significantly weaker middle trapezius, lower trapezius, and serratus anterior (p<0.05). Both CVA and RSA showed negative correlations with protraction, anterior tilt of the scapula; they showed moderate positive correlations with scapular elevation. The strengths of the middle trapezius, lower trapezius, and serratus anterior muscles were moderately correlated with CVA and RSA (p<0.05). Conclusions Relative to normal subjects, those with FHRSP exhibited changes in scapular position and periscapular muscle deficits. This information will be useful for designing exercise programs for individuals with FHRSP.

Infectious disease spread is a multiscale process composed of within-host (biological) and between-host (social) drivers and disentangling them from each other is a central challenge in epidemiology. Here, we introduce VIBES, a multiscale modeling framework that explicitly integrates viral dynamics based on patient-level data with population-level transmission on a data-driven network of social contacts. Using SARS-CoV-2 as a case study, we analyze three emergent epidemic properties, namely the generation time, serial interval, and presymptomatic transmission. First, we established a purely biological baseline, thus independent of the reproduction number (R), from the within-host model, estimating a generation time of 6.3 d for symptomatic individuals and 43.1% presymptomatic transmission. Then, using the full model incorporating social contacts, we found a shorter generation time (5.4 d at R = 3.0) and an increase in presymptomatic transmission (52.8% at R = 3.0), disentangling the impact of social drivers from a purely biological baseline. We further show that as pathogen transmissibility increases (R from 1.3 to 6), competition among infectious individuals shortens the generation time and serial interval by up to 21% and 13%, respectively. Conversely, a social intervention, like isolation, increases the proportion of presymptomatic transmission by about 30%. Our framework also estimates metrics that are challenging to obtain empirically, such as the generation time for asymptomatic individuals (5.6 d; 95%CI: 5.1 to 6.0 at R = 1.3). Our findings establish multiscale modeling as a powerful tool for mechanistically quantifying how pathogen biology and human social behavior shape epidemic dynamics as well as for assessing public health interventions.