Detecting cyber security related Twitter accounts and different sub-groups

Abstract

their generalizability and validation of performance. In addition, there is a lack of more general-purpose sub-classifiers that can classify different sub-groups of cyber security related accounts, e.g., cyber security individuals (vs. groups and organizations), hackers in general (both people and groups), researchers and research organizations, etc. Such sub-classifiers will allow more fine-grained monitoring of the different sub-groups to support more targeted monitoring and behavioral analysis.In this paper, we report our work that addresses a number of the issues about classifying cyber security related accounts on Twitter. Our work is based on a three-staged methodology: a more systematic data collection process, crowdsourcing-based labeling experiment, and development of machine learning based classifiers. Our main contributions are as follows:

Open Source Intelligence Opportunities and Challenges: a Review

“…Similar research was provided by Mahaini et al in [70]. The authors focused on detecting cybersecurity-related Twitter accounts and different sub-groups.…”

“…These classifiers include a baseline classifier for identifying accounts related to cybersecurity generally and three sub-classifiers for identifying accounts related to individuals, hackers. Nobili et al in [70] focused on European violence against workers. They noticed that this problem requires a combination of safety and security perspectives.…”

“…Penetration testing automation [66] Characterization and profiling the identified threats [67] Detecting cybersecurity-related Twitter accounts and different sub-groups [70] Compile data regarding the Pulwama attacks [74] Phishing prevention [79] Social Media group solutions. The first group was concerned with detecting and solving cyberspace problems to increase Internet users' security.…”

CybAttT: A Dataset of Cyberattack News Tweets for Enhanced Threat Intelligence

“…Another recent work by Mahaini et al [23] produced several machine learningbased classification models to detect cybersecurity-related accounts on X. The X platform Sampling Application Programming Interface (API) was used to collect tweets that include cybersecurity related discussions.…”

“…It can be noticed that some of the reviewed machine learning and deep learning based models have been applied to datasets including small numbers of collected tweets, such as the models proposed by Deshmukh et al [24] and Behzadan et al [25]. Moreover, the reviewed research studies have focused on applying either machine learning algorithms, as the models proposed by Le et al [20], Ghankutkar et al [21], Arora et al [22], Mahaini [23], and Deshmukh et al [24], or deep learning algorithms, as the models proposed by Behzadan et al, Behzadan, Dionísio et al [26], and Coyac-Torres et al [27]. There are no comparisons conducted between both learning types to discover the most adequate learning algorithm for the proposed cybersecurity data classification models.…”