Enhancing topic clustering for Arabic security news based on k‐means and topic modelling

Abstract

The internet has become one of the main sources of news spread as it unleashed the information dissemination space, where the news websites express opinions on entities while also reporting on recent or unusual security risks. Recently, many research studies have focused on sentimental reflection on the views and impressions of people utilising natural language processing and analytical linguistics. Therefore, we have collected corpus from popular Arabic websites that publish articles related to recent security issues, and we provide light weight preprocessing techniques where data is term matrix is transformed. We also present an intensive lexical-driven data analysis with visualised data views, as our topic modelling technique can effectively extract significant topics from all the collected text from different websites. Our experiments validate the k-means clustering algorithm with and without the latent Dirichlet allocation topic modelling method, and we adopted various validation techniques to measure the topic clustering internally and externally. As shown in the experiments' results, our proposed combined method has a high round index rate of 87.2%, with a large number of topics and clusters. This is an open access article under the terms of the Creative Commons Attribution-NonCommercial-NoDerivs License, which permits use and distribution in any medium, provided the original work is properly cited, the use is non-commercial and no modifications or adaptations are made.

Codeguard: Utilizing Advanced Pattern Recognition in Language Models for Software Vulnerability Analysis

“…Static analysis methods, which extract key features from code for input into machine learning models, have been a focal point of some studies (Kim et al, 2022;Li et al, 2016. In contrast, others have utilized dynamic analysis, where the code is executed, and its behavior monitored to identify vulnerabilities (Alharbi, Hijji & Aljaedi, 2021;Salimi & Kharrazi, 2022).…”

“…Deep learning architectures such as Convolutional Neural Networks (CNNs) and RNNs have been extensively studied for vulnerability detection (Alharbi, Hijji & Aljaedi, 2021;Kim, Woo, Lee & Oh, 2017;Kim et al, 2022;Li et al, 2016;Rabheru, Hanif & Maffeis, 2021;Salimi & Kharrazi, 2022;Yamaguchi, Golde, Arp & Rieck, 2014;Zhou & Verma, 2022). These models typically require structured data for identifying features linked to vulnerabilities.…”

SecuGuard: Leveraging pattern-exploiting training in language models for advanced software vulnerability detection

“…Myriad studies have unveiled methodologies that employ machine learning techniques for this pursuit. Some concentrate on static analysis, extracting salient features from the code to be input into predictive machine-learning frameworks [1,2], while others hinge on dynamic analysis, running the code and monitoring its behavior to discern vulnerabilities [3,4]. A nascent inclination towards harnessing deep learning models for source code vulnerability detection has been observed.…”

“…Meanwhile, other investigations employ transformers, which have gained acclaim in the realm of natural language processing [9,10]. Literature has extensively analyzed deep learning architectures like Convolutional Neural Networks (CNNs) and Recurrent Neural Networks (RNNs) for vulnerability detection [3,[12][13][14]25,29,31,34]. Yet, these models necessitate structured data to discern vulnerability-associated features.…”

Topic Modeling: Perspectives From a Literature Review

“…Finally, it is worth highlighting that a series of studies has considered the presence of topics over time. These studies identified and evaluated expert opinions on COVID -19 [72] during the outbreak period compared with other periods [73].…”