Table 1.

Previous findings in the field.

Study	Aim	Methods	Data	Main results
Email classification research trends: review and open issues (Mujtaba et al., 2017)	Review of e-mail classification methods from 2006 to 2016, analyzing five aspects: application areas, datasets, feature spaces, classification techniques, and performance measures	Comprehensive review and analysis	98 articles (56 articles from Web of Science core collection databases and 42 articles from Scopus database)	The authors identify five techniques—supervised, semi-supervised, unsupervised, content-based, and statistical learning—with supervised ML being the most common and Support Vector Machine showing the best performance, followed by Decision Trees and Naive Bayes
A recent review of conventional vs. automated cybersecurity anti-phishing techniques (Qabajeh et al., 2018)	Examination of the effectiveness of traditional anti-phishing approaches, such as awareness campaigns, user education, and periodic training sessions, in comparison to computerized anti-phishing techniques	Classification of anti-phishing approaches in the analyzed literature into 3 main categories: “education and legal, computerized using human-crafted methods, and intelligent ML methods”	75 studies	ML and rule induction are particularly effective in phishing prevention, offering high detection accuracy and easily interpretable results. The tendency to use ML and DL algorithms for website classification to identify phishing sites was considered promising by the authors for reasons of cost and accuracy
Evaluation of phishing techniques based on machine learning (Kunju et al., 2019)	Survey of phishing attacks and their detection methods, with the intention to raise user awareness about the associated risks, and present various machine learning techniques (kNN, Naïve Bayes, Decision Tree, SVM, Neural Network, Random Forest) used for predicting and preventing phishing websites	Overview of ML algorithms for detecting phishing websites, including k-Nearest Neighbors (kNN), Naïve Bayes, Decision Trees, Support Vector Machines (SVM), Neural Networks, and Random Forest	14 studies	The necessity of employing multiple techniques to enhance phishing detection effectiveness is highlighted
Toward the detection of phishing attacks (Athulya and Praveen, 2020)	The paper aims to raise user awareness about phishing strategies and present a hybrid detection method that offers fast response time and high accuracy	Review of various phishing attacks, evasion techniques, and anti-phishing approaches	9 research articles	The most effective approach to mitigating phishing attacks is raising user awareness and selecting the most appropriate anti-phishing security software
Toward a systematic description of the field using bibliometric analysis: Malware evolution (Mat et al., 2021)	A bibliometric analysis of a decade of evolution in malware research, considering it as an umbrella term for all malicious software, with a specific focus on Android malware due to a significant rise in occurrences in 2019	Bibliometric review	1,278 articles	The article does not explicitly address phishing
Applications of deep learning for phishing detection: A systematic literature review (Catal et al., 2022)	Analysis of the use of DL for phishing detection	Systematic literature review	43 studies	The most commonly used algorithm is the Deep Neural Network (DNN), followed by Convolutional Neural Networks (CNN) and Recurrent Neural Networks (RNN)/Long Short-Term Memory Networks (LSTM). The study also indicates that DNN and Hybrid DL algorithms achieved the best performance in phishing detection
A bibliometric analysis of phishing in the Big Data era: high focus on algorithms and low focus on people (Pejić-Bach et al., 2023)	A co-occurrence analysis using VOSviewer on a set of 136 articles focused on big data and phishing	Bibliometric review	NA (WoS database)	Predominantly technical research (computer science, engineering, telecommunications); big data ML cluster emphasizes ML/DL benefits for real-time anti-phishing; approaches include models, voting frameworks, consensus clustering, URL analysis; Gray Wolf Optimizer outperforms other algorithms via feature analysis (e.g., URL length, HTTP response)
A systematic literature review on phishing website detection techniques (Safi and Singh, 2023)	An update in the previous systematic literature surveys with more focus on the latest trends in phishing detection techniques	Systematic literature review	80 scientific papers published between 2017 and 2021	ML techniques dominated phishing detection (71.25%), followed by heuristic (66.25%), visual similarity (43.75%), DL-based (17.5%), and list-based methods (12.5%); PhishTank was the main data source, while Random Forest, SVM, and Decision Tree were the most used ML algorithms, with CNN achieving the highest accuracy (99.98%)
Mapping the phishing attacks research landscape: a bibliometric analysis and taxonomy (Mutluturk and Metin, 2023)	A holistic approach of the topic and presentation of an in-depth analysis of phishing research from 2004 to 2023, emphasizing the field’s steady growth, emerging trends, and collaborative networks	Bibliometric review	3,139 phishing-related articles indexed in the Web of Science database	ML-based techniques play a central role in phishing research, with CANTINA+ (Xiang et al., 2011) ranking 3rd among the Top 10 Most Cited Publications (2004–2013), after studies by Jagatic et al. (2007) and one on the economics of information security (Science). In 2014–2023, Sahingoz et al. (2019) ranks 2nd, followed by Chiew et al. (2019) and Mahdavifar and Ghorbani (2019), highlighting the growing impact of AI-related approaches (Mutluturk and Metin, 2023). Zhang and Xiang emerge as key co-citation nodes, while Chiew leads another cluster; Cranor and Hong are the most cited authors. Among keywords, Machine Learning ranks 3rd and Neural Networks 12th
Enhancing spear phishing defense with AI: a comprehensive review and future directions (Mohamed et al., 2024)	A critical analysis of AI techniques, including ML, NLP, but also behavioral analytics, mitigating spear phishing attacks	Comprehensive review	30 seminal papers	ML models are effective for pattern recognition but require extensive training data, whereas NLP techniques enhance contextual and semantic understanding, improving detection of sophisticated phishing attempts