2022 IEEE International Conference on Computing (ICOCO) 2022 IEEE International Conference on Computing (ICOCO) | 978-1-6654-8996-6/22/$31.00 ©2022 IEEE | DOI: 10.1109/ICOCO56118.2022.10031671 A Comparative Study on Email Phishing Detection Using Machine Learning Techniques Afiqah Aqilah Adzhar Faculty of Computer and Mathematical Sciences Universiti Teknologi MARA 40450 Shah Alam, Selangor, Malaysia afiqah.adzhar@gmail.com *Zulaile Mabni Faculty of Computer and Mathematical Sciences Universiti Teknologi MARA 40450 Shah Alam, Selangor, Malaysia zulaile@fskm.uitm.edu.my Zaidah Ibrahim Faculty of Computer and Mathematical Sciences Universiti Teknologi MARA 40450 Shah Alam, Selangor, Malaysia zaidah@fskm.uitm.edu.my In this paper, Phishing Email, Phishing Attack, and the previous studies of different supervised machine learning techniques in detecting Phishing Email are discussed. The overall focus of this paper is to find out which supervised machine learning techniques is the best technique in detecting phishing email. Abstract— Phishing Email can be described as an email that looks exactly like a legitimate email, but it is designed by phisher with an intention to deceive the email’s user. The purpose of phishing email is to trick email user to visit fake website that looks exactly like a real one or to trick user to download the available attachment in the email without knowing that they are downloading virus into their machine. As the number of phishing emails are increasing from day to day and due to the complexity in detecting phishing email, there are numbers of continuous researches that have been done to improve existing detection tools or to develop a new one. To provide a thorough understanding of phishing attacks, this paper provides a brief explanation on phishing email and phishing attack. This paper presents the comparison of previous studies in commonly used Supervised Machine Learning techniques on detecting the phishing email attack such as Decision Tree (DT), Naive Bayes (NB), Random Forest (RF), and Support Vector machine(SVM). The findings of this study concluded that SVM and RF are the best techniques that can be used to detect phishing email. II. PHISHING EMAIL AND PHISHING ATTACKS Phishing email is known as one of the most famous cyber-attacks. It can be described as an email that looks exactly like a legitimate email, but it is designed to manipulate email user to get their sensitive information or to install malware into their machine [8]. This activity is said to be successfully done if user click on the links and key in their credentials or download any document provided in the email. According to [3], phishing can be done traditionally and in real time. Fig. 1 shows an illustration on the steps in a traditional asynchronous phishing attack [3]. It starts with victim submits their credential to a phishing website. Then, the phishing site will collect all the sensitive information and phisher will initiate to login in real websites using the stolen credentials. Fig. 2 illustrate on how phishing was done in real time [3]. It starts with human activity where victim submits their credential to RTPP page then RTTP will collect the victim’s credential in real time. At the same time RTTP will initiate an automated login attempt and phisher also can manually initiate login attempt using the stolen credentials. Keywords—Phishing, Email, Machine Learning, Decision Tree, Naïve Bayes, Random Forest, Support Vector Machine I. INTRODUCTION Phishing is a famous type of social engineering attacks that take human good nature for granted [1]. Phishing can be described as a method where the phisher created some fake websites that looks exactly like the original one [2]. According to research, 96% of phishing attack is coming from email, while 3% is from harmful websites and another 1 % is through phone [3]. Email or also known as electronic mail is one of the common and efficient communication methods. It is used to exchange messages between two or more users for formal and informal activities [4]. According to the Statista research, approximately 4 billion email users are recorded in 2020 and expected to grow up to 4.6 billion users in 2025 [3]. The increasing number of email user has brought them to a wide range of security risk [5]. One of the security risks that they are facing is phishing email. According to [6], phishing email can be described as a situation where attacker send an email that trick user to visit fake website that looks exactly like a real one. The purpose of email phishing is to gain access to sensitive information data or to deploy malicious software to user’s computer [7]. Fig 1: Steps in Traditional Asynchronous Phishing Attack Authorized licensed use limited to: The University of West London. Downloaded on January 08,2025 at 17:14:58 UTC from IEEE Xplore. Restrictions apply. 978-1-6654-8996-6/22/$31.00 ©2022 IEEE 96 2022 IEEE International Conference on Computing (ICOCO) The first group is known as Link-Based Group where link is provided in the email [1]. Next is Text-Based Group [1]. It is an attack that depends only on recipients psyche to be attracted so that they will give away the password. Third one is Image-Based Group [1]. This group is quite difficult to detect as it is using a combination of text, image, and link to trick the user to visit the phishing website. Fig 2: Real Time Phishing Proxies (RTTP) Reusing a Victim’s Data in Real Time A. Characteristic of Phishing Email Phishing email can be easily traced if we can identify its characteristic in an early stage. There are five common characteristic of Phishing email [9]. The first one is too good to be true [9]. It is a situation where phisher will design an email that will use catchy and interesting statement to grab people’s attention. The statements could be some urgent statement or an exciting statement. Lastly is Attachment-Based Group [1]. It is an email with attachment that containing malicious software. The attachment appeared in the form of benign looking HTML files, documents, or PDF attachment. User that failed to identify this email as phishing email will download the attachment and ended up downloading malicious software to their machine. III. DETECTION TECHNIQUES Machine Learning (ML) is an iterative process. Machine Learning models are taught from previous experiences as well as historical data analysis. On top of that, ML models can recognize trends to generate predictions about the dataset's future [10]. The process of ML can be divided into 5 stages as illustrated in Fig. 3. Next is sense of urgency [9]. Phisher will design an email that asking user to act fast as the deals are limited or user’s account is about to expire, or account is suspended, and they only have short time to respond. The third characteristic is Hyperlink [9]. Hyperlink is used to display a link and upon clicking on the hyperlink it will redirect user to actual site. In common case of email phishing, phisher will display a legit hyperlink or a button and hide the actual link to deceive users. The fourth characteristic is attachment [9]. Phisher will attach the email with some file. The attachment might be also named with something catchy. This file might contain ransomware or any other virus that when user download the attachment, the viruses or ransomware will also be downloaded to user’s machine. Lastly is unusual sender [9]. Phishing email could be coming from someone that the user knows or someone they didn’t know. As it might also coming from someone user know, phishing email is always unexpected and suspicious one. B. Groups of Phishing Attacks Phishing attack can be defined as a situation where victim is being directed to a fake website that looks exactly like an original website [2]. A study by [1], states that there are 4 groups of phishing attacks. Fig 3: Stages of ML process The first step is Data Collections. The data set can be collected from a variety of sources, including a file, a database, a sensor, online and many other places. Although it can be collected easily but it cannot be utilized immediately for analysis since there may be a lot of missing data, extremely big values, disorganized text data, or noisy data. Therefore, Data Preparation is completed to address this issue. Next is data preparation. Data Preparation is the most important process as it helps in building a ML model with high accuracy. This step starts with data pre-processing in which it is the process of cleaning the raw data set to have a small clean data set [10]. Next step is to build a model for selecting correct ML Algorithm. The major goal of this step is to use the preprocessed data to train the best performing model feasible. There are two types of learning in machine learning. The Authorized licensed use limited to: The University of West London. Downloaded onfirst January at 17:14:58 ML UTC from Xplore. Restrictions apply. one08,2025 is Supervised and IEEE another one is Unsupervised 97 2022 IEEE International Conference on Computing (ICOCO) Machine Learning. This paper will focus on Supervised Machine Learning. The fourth step is to evaluate the model. In order to evaluate the classification model there are some of best and most used evaluation methodology that can be use such as Confusion Matrix, Accuracy and also AUC and ROC [10]. Last step is to deploy the ML model into production environment. This implies integrating the completed model into a production environment and obtaining results to make business decisions [10]. A. Supervised Machine Learning Supervised learning is a type of machine learning that use labeled data to train algorithm to classify the data or to predict the outcomes of the data [11]. Supervised learning is beneficial to organization as it helps to solve variety of realworld problem. In supervised machine learning, there are two categories of algorithms which are classification and regression. The output of classification is a choice between classes such as True or False. Meanwhile for regression, the output will be in the form of real numbers. For example, regression is used to predict the stock prices. The example of approaches and algorithm used for supervised machine learning are Naive Bayes classifiers, Decision Trees (DT), Support Vector Machines (SVMs) and Random Forests (RF). 1) Naïve Bayes (NB) Naive Bayes is one of the classification techniques that is based on Bayes’ theorem [12]. The formula of Bayes’ Theorem is as shown in (1) below. each feature has the same of level of importance. Therefore, it cannot learn the relationship between features [13]. There are three types of Naïve Bayes models which are known as Gaussian, Multinomial and Bernoulli [13]. The Gaussian model implies that characteristics are distributed normally. If predictors accept continuous values rather than discrete values, the model assumes that these values are drawn from a Gaussian distribution. When the data is multinomial distributed, the Multinomial Nave Bayes classifier is utilized. It is generally used to solve document categorization issues, which involves determining which category a document belongs to. The predictions in the classifier are based on the frequency ofterms. The Bernoulli classifier operates in a similar way to the Multinomial classifier, except that the predictor variables are independent Boolean’s variables. For example, determining whether a specific word appears in a document. This paradigm is also well- known for jobs involving document categorization. 2) Support Vector Machine (SVM) SVM is one of the computer algorithms that works by learning from example to assign label to projects. It is very important to grasp four basic concepts to understand the essence of SVM classification. The four basic concepts are [14]: (i) the separating hyperplane (ii) the maximum-margin hyperplane (iii) the soft margin and Where: (iv) the kernel functions. P(A|B) – the probability of event A occurring, given event B has occurred P(B|A) – the probability of event B occurring, given event A has occurred P(A) – the probability of event A P(B) – the probability of event B The objective of SVM is to identify the hyperplane in the number of features-dimensional space which classifies the data points distinctly. As shown in Fig. 4 is the example of possible hyperplane. P B|A . P(A) P(A|B)= P(B) (1) There are three steps in this technique. Firstly, convert the data set into frequency table. Then create a table of likelihoods by calculating the probabilities. Lastly, compute the posterior probability using the Naive Bayesian equation for each class. The outcome of prediction is the class with the highest posterior probability [12]. One of the fast and easy ML algorithms to predict a class of datasets is Naïve Bayes algorithm [13]. It can be used for Binary as well as multi-class classification. In fact, compared to other algorithms, Naïve Bayes performs well in multi-class predictions. Lastly, for text classification issues, the most used method is Naive Bayes [12]. Fig 4: Possible Hyperplane By referring to Fig. 4, to separate two classes of data points, many hyperplanes can be selected but the objective here is to find a plane with a maximum margin, for example the maximum distance between data points of both classes [15]. Hyperplanes or also known as decision boundaries are important as it helps to helps to classify data points. Any data points that fall at either side of the hyperplane may be classified as members of different classes. The only disadvantage in Naive Bayes technique is that it assumes the occurrence of one feature in a class has no bearing on the presence subsequent and Authorized licensed use limited to: Theof University of West features London. Downloaded on January 08,2025 at 17:14:58 UTC from IEEE Xplore. Restrictions apply. 98 2022 IEEE International Conference on Computing (ICOCO) 3) Decision Tree (DT) Decision Tree are mainly used for solving classification problems, but it can be used for both classification and regression problems [16]. It is a tree-structured classifier where internal nodes contain dataset attributes, branches represent decision rules, and each leaf node provides the end result. Decision Tree works in a way where the tests are carried out based on the characteristics of the provided dataset. It is called decision tree as it has similar characteristics of a tree in which it starts with the root node, and then spreads out to form a tree-like structure. In order to form a tree, Classification and Regression Tree (CART) algorithm will be utilized. Decision tree works by simply asks a question and based on the answer which is between Yes or No or any numeric data, then it will further split the trees into subtrees. Fig. 5 below shows the illustration on the general structure of a decision tree. Some disadvantages of decision tree are it is complex as it contains a lot of layers. Apart from that, it may have an overfitting problem, which can be fixed by the Random Forest algorithm [16]. 4) Random Forest (RF) Random Forest algorithm is known as an easy to use and flexible machine learning algorithm. Random Forest is also one of the most used machine learning algorithms as it can be used for both classification and regression task. It consists of large number of an individual decision trees that works as an ensemble. As shown in Fig.6, the final model’s prediction is depending on individual tree as each of it will determine their class prediction either 1 or 0 and the highest number of class prediction will be the final model’s prediction. For example, as seen in Fig.6, there are a total of 9 individual trees and there are six trees that vote for 1 and 3 votes for 0, therefore the model’s prediction will be 1 as it has the highest number of votes [17]. Fig 5: Illustration on general structure of a tree Fig.5 shows that the decision tree starts with Root Node and ends with Leaf Node. Root node represents the full dataset, which it is then split into two or more homogenous groups. Meanwhile leaf node is the last output nodes in which the tree cannot be further segmented once a leaf node has been obtained. There are two processes involved in decision tree which are Splitting and Pruning [16]. Pruning is the process where it will remove all the unwanted branches from the tree. While the process of dividing the decision node/root into sub-nodes based on the given condition is called splitting. The outcome of splitting process is called branch or subtree in which the tree is formed. In a decision tree, the process of determining class of a given dataset starts at the root node of the tree. This algorithm checks the values of the root property with the values of the record (actual dataset) attribute and then follows the branch and jumps to the next node depending on the comparison. Next, the attribute value is checked with the other sub-nodes and moves on to the next node. The process is repeated until it reaches the tree's leaf node. Fig 6: Visualization of Random Forest Model Making Prediction There are prerequisites to make sure the Random Forest to perform well which are [17]: (i) Some actual signal is needed in our features so that models built using those features do better than random guessing. (ii) Low correlation of the predictions (and therefore the errors) made by the individual trees. B. Unsupervised Machine Learning Unsupervised learning occurs when an AI system is given unlabeled, uncategorized data and its algorithms act on it without any prior training. Unsupervised ML can be categorized into two categories which are Clustering and Association. IV. RESULTS AND ANALYSIS There are few advantages of Decision Tree such as it is a simple and easy to understand as it follows the same This section evaluates the previous studies on phishing human decision-making process in real-life. Next, it is detection. Most of the works includes the use of Machine useful in solving decision-related problems as it helps to Learning and Deep Learning techniques. Table I shows the think of all the possible outcomes for a problem. Lastly results of the comparison and literature review. compared to other methods, data cleansing is not required as Authorized much in licensed Decision algorithm[16]. usetree limited to: The University of West London. Downloaded on January 08,2025 at 17:14:58 UTC from IEEE Xplore. Restrictions apply. 99 2022 IEEE International Conference on Computing (ICOCO) Basit et al. has done a literature review on Artificial Intelligence techniques where the authors are comparing the available techniques of Machine Learning, Deep Learning, Hybrid Learning as well as Scenarios based techniques for Phishing detection. The methods that they are comparing are SVM, RF, ANN, C 4.5 Algorithm, k-NN and DT. After a detailed comparative study, researcher found out that machine learning methods are the best methods in detection. This is because RF outperforms all other classification methods in terms of performance and accuracy as tested for different datasets [18]. Authors in [19] used deep semantic analysis, machine learning, and deep learning approaches to extract inherent properties of email text and identify emails as phishing or non- phishing in another experiment. Their findings show that Deep Learning models performed a little better than the Machine Learning models. On the other hand, the accuracy achieved by SVM was slight better with word phrasing is 93.88% then without word phrasing is 82.35%. An experiment that uses novel approach for Phishing URLs detection using lexical based machine learning in real-time has been conducted [23]. In this research authors focuses on conducting the experiment using RF, KNN, SVM and also Logistic Regression techniques. The selected features in this experiment are chosen on lexical based. Random Forest has the highest accuracy of 99.57%. Lastly is an experiment conducted by Sahingoz et al. This experiment focuses on machine learning based detection on URLs [2]. The authors conducted the experiment on 7 different techniques which are DT, Adaboost, Kstar, KNN, Random Forest (RF), SMO, Naïve Bayes (NB) and the features are selected based on NLP features, Word Vector as well as Hybrid. Random Forest with NLP features has the highest accuracy which 97.98% compared to others. TABLE I. Comparisons Between Supervised ML Techniques Another experiment was done by Rahaf et al.,where Support Vector machine (SVM), Random Forest (RF), Naïve Bayes (NB), Decision Tree (DT) algorithms were tested. The results are generated and compared to determine the most effective algorithm for fraud detection. Their findings show that SVM has the best performance as it has achieved accuracy of more than 98% [20]. AUTHORS TECHNIQUES USED RESULTS Basit et. al SVM , RF, ANN , C4.5 Algo, k-NN and DT NB, SVM, DT, LSTM, CNN, Word Embedding Several studies proved that more than 95% attack detection accuracy can be achieved using a RF classification method. SVM , RF, NB , DT DT, KNN , NB, RF, SVM and Logistic Regression SVM has the best performance as it has achieved accuracy of more than 98% Sikha et al Another experiment was conducted by Anu et al.to model phishing email detection as a supervised classification problem to detect phishing emails from legitimate ones. In this experiment they used nonsequential representation such as Term Document Matrix (TDM) approach followed by Singular Value Decomposition (SVD) and Nonnegative Matrix Factorization (NMF). The machine learning techniques that the authors used in this experiment are Decision Tree (DT), K- Nearest Neighbors (KNN), Logistic Regression, Naive Bayes (NB), Random Forest (RF), and SVM. From their findings, it shows that Random Forest has outperformed all other techniques for the training data set with accuracy of 88.906% for TDM with SVD representation and 89.06 % for TDM with NMF representation [21]. Rahaf al et. Anu et. al Deep Learning models performed a little better than the ML models. However the accuracy of SVM was slight better with word phrasing (93.88% ) then without word phrasing (82.35%). Random Forest has outperformed all other techniques for the training data set. Accuracy: TDM with SVD representation is 88.906% TDM with NMF representation is 89.06 % Another research focusing on implementing a system that able to distinguish between legitimate and malicious emails using supervised machine learning. The techniques that the authors have implemented in their system are Random Forest (RF), Decision Tree (DT), Supervised Machine Learning (SVM) and K- Nearest Neighbors (KNN). From their experiment they found out that Random Forest is the most suitable approach for email classification and authors also able to identify 4 different categories in which this algorithm can include a new email which are: Spam, Scam, Phishing and legitimate emails [22]. Eva et al RF, DT, SVM and KNN Random Forest is the most suitable approach for email classification Paliathet. al KNN, SVM, NB, Rough Set, RF and Random Tree Precision NN: 99.76% SVM: 99.75%. Gupta et al Next, in order to find the most effective phishing detection using selected features authors applied Machine Learning approach in their research such as Neural Network (NN), SVM, NB, Rough Set, RF and also Random Tree. The experiment took place based on 8 selected features that is based on information Gain (IG) Calculation [1]. Neural Network and SVM has the highest precision of 99.76% and 99.75% respectively. NN has the highest percentage of accuracy that is up to 99.44%. Sahingoz et al RF, KNN , SVM , Logistic Regression techniques DT, Adaboost, Kstar, kNN, RF, SMO, NB Accuracy: NN has the highest percentage of accuracy that is up to 99.44% Accuracy: Random Forest 99.57% KNN 99.04% SVM 97.64% Logistic regression 95.56% Accuracy: Random Forest with NLP features has the highest accuracy which 97.98% compared to others. Authorized licensed use limited to: The University of West London. Downloaded on January 08,2025 at 17:14:58 UTC from IEEE Xplore. Restrictions apply. 100 2022 IEEE International Conference on Computing (ICOCO) V. CONCLUSION Number of phishing email has been increasing from day to day. From this paper we have described what phishing email is, the characteristics of phishing email and group of phishing attacks. Apart from that, we have also identified what is Machine Learning, type of Machine Learning and described the process of the Machine Learning as well as the few types of Supervised Machine Learning that can be used for detection such as Naïve Bayes, Decision Tree, Support Vector Machine and Random Forest. A comparative study on the mentioned supervised machine learning techniques has been discussed. From the study, we can conclude that SVM and RF are the best techniques that can be used to detect phishing email. ACKNOWLEDGMENT The registration fee is funded by Pembiayaan Yuran Prosiding Berindeks (PYPB), Tabung Dana Kecemerlangan Pendidikan (DKP), Universiti Teknologi MARA (UiTM), Malaysia. REFERENCES [1] Research in Business & Social Sciences, 2021, vol. 11(12), pp. 2633 – 2643. [12] S. Ray, Learn Naive Bayes Algorithm | Naive Bayes Classifier Examples. [online] Analytics Vidhya. Available at:<https://www.analyticsvidhya.com/blog/2017/09/naive-bayesexplained/> [Accessed 1 February 2022]. [13] www.javatpoint.com. n.d. Naive Bayes Classifier in Machine Learning Javatpoint. [online] Available at:<https://www.javatpoint.com/machine-learning-naive-bayesclassifier> [Accessed 1 February 2022]. [14] W. S. Noble, What is a support vector machine? Nature Biotechnology,2006, http://www.nature.com/naturebiotechnology [15] Support Vector Machine — Introduction to Machine Learning Algorithms|by Rohith Gandhi|Towards Data Science.(n.d.). Retrieved July 10, 2021, From: https://towardsdatascience.com/support-vector-machine introduction-to-machine-learning-algorithms-934a444fca47 [16] www.javatpoint.com. n.d. Machine Learning Decision Tree Classification Algorithm - Javatpoint. [online] Available at: <https://www.javatpoint.com/machine-learning-decision-treeclassification-algorithm> [Accessed 2 February 2022]. [17] T. You, Understanding Random Forest. How the Algorithm Works and Why it is So Effective, Towards Data Science. (n.d.). 2019, Retrieved July 23, 2021 [18] A. Basit, M. Zafar, X. Liu, A. R. Javed, Z. Jalil, & K. Kifayat, A comprehensive survey of AI-enabled phishing attacks detection techniques. Telecommunication Systems, 2021, vol. 76(1), pp. 139154. [19] S. Bagui, D. Nandi, S. Bagui, &R. J. White, Classifying phishing email using machine learning and deep learning. In 2019 International Conference on Cyber Security and Protection of Digital Services (Cyber Security), 2019, pp. 1-2. IEEE. [20] R. Al-Haddad, F. Sahwan, A. Aboalmakarem, G. Latif, &Y. M. Alufaisan, Email text analysis for fraud detection through machine learning techniques. In 3rd Smart Cities Symposium (SCS 2020), 2020, pp. 613-616, IET. [21] A. Vazhayil, N. B. Harikrishnan, R. Vinayakumar, K. P. Soman,&A. D. R. Verma, PED-ML: Phishing email detection using classical machine learning techniques. In Proc. 1st Anti Phishing Shared Pilot at 4th ACM International Workshop on Security & Privacy Analytics (IWSPA 2018), pp. 1-8, Tempe, AZ, USA. [22] E. Marková, T. Bajtoš, P. Sokol and T. Mézešová, "Classification of malicious emails," 2019 IEEE 15th International Scientific Conference on Informatics, 2019, doi: 10.1109/Informatics47936.2019.9119329 [23] B. B. Gupta, K. Yadav, I. Razzak, K.Psannis, A. Castiglione, &X. Chang, A novel approach for phishing URLs detection using lexical based machine learning in a real-time environment. Computer Communications, 2021, vol. 175, pp. 47-57. https://doi.org/10.1016/J.COMCOM.2021.04.0 S. Paliath, M. A. Qbeitah, & M. Aldwairi, Phishout: Effective phishing detection using selected features. Proceedings of the 2020 27th International Conference on Telecommunications,ICT2020. https://doi.org/10.1109/ICT49546.2020.9239589 [2] O. K. Sahingoz, E. Buber, O. Demir, & B. Diri, Machine learning based phishing detection from URLs. Expert Systems with Applications, 2019, vol. 117,pp. 345–357. https://doi.org/10.1016/J.ESWA.2018.09.029 [3] 2020 Phishing and Fraud Report. (n.d.). Retrieved July 23, 2021, from https://www.f5.com/labs/articles/threat intelligence/2020phishing-and-fraud-report [4] Y. A. Yaseen, M. Qasaimeh, R. S. Al-Qassas, &M, Al-Fayoumi, Email fraud attack detection using hybrid Machine Learning approach. Recent Patents on Computer Science, J u n e 2 0 1 9 , v o l . 12 (5). https://doi.org/10.2174/2213275912666190617162707 [5] X. Li, D. Zhang, & B. Wu, Detection method of phishing email based on persuasion principle. Proceedings of 2020 IEEE 4th Information Technology, Networking, Electronic and Automation Control Conference, ITNEC2020, pp. 571–574. https://doi.org/10.1109/ITNEC48623.2020.9084766 [6] R. A. A. Helmi, C. S. Ren, A. Jamal & M. I. Abdullah, Email antiphishing detection application. 2019 IEEE 9th International Conference on System Engineering and Technology, ICSET2019Proceeding,pp. 264–267. https://doi.org/10.1109/ICSENGT.2019.8906316 [7] N. Moradpoor, B. Clavie, &B. Buchanan, Employing machine learning techniques for detection and classification of phishingemails.ProceedingsofComputingConference2017,January 2018, pp. 149–156. https://doi.org/10.1109/SAI.2017.8252096 [8] M.H. Ramli, A. I. H. Faugi, N. M. Mohd Faizal, N. Mohd Khadri, J. Mohamad Zain, Anti-phishing with google extension 3H1M using blacklist algorithm, Malaysian Journal of Computing, vol. 5(1),pp. 362-373, 2020 [9] Phishing | What Is Phishing? (n.d.). Retrieved July 5, 2021, from https://www.phishing.org/what-is-phishing [10] S. Pandian, Understand Machine Learning | What is Machine Learning ?. [online] Analytics Vidhya. Available at: <https://www.analyticsvidhya.com/blog/2020/12/understandmachine-learning-and-its-end-to-end-process/> [Accessed 31 January 2022]. [11] Z. Ibrahim, N. Mat Diah, N. A. Rizal, &M. N. Yuri, Prediction of early symptoms of COVID-19 infected patients using Supervised Machinelicensed Learning models,to: The International Journal of Academic Authorized use limited University of West London. Downloaded on January 08,2025 at 17:14:58 UTC from IEEE Xplore. Restrictions apply. 101
0
You can add this document to your study collection(s)
Sign in Available only to authorized usersYou can add this document to your saved list
Sign in Available only to authorized users(For complaints, use another form )