Web2Vec: Phishing Webpage Detection Method Based on Multidimensional Features Driven by Deep Learning

Zou, Lianyang; Ye, Ou; Han, Jingzhou

doi:10.1109/access.2020.3043188

Cited by 55 publications

(72 citation statements)

References 21 publications

(55 reference statements)

Supporting

Mentioning

Contrasting

Order By: Relevance

“…CNN was used as a single classifier in numerous research to distinguish between phishing and legitimate websites [7,8,20,[24][25][26][27][28]. It can also be used in combination with other DL techniques to form an ensemble model and to improve phishing detection accuracy [10,11,[29][30][31][32][33][34][35][36]. The difference between the architectures of CNN and DNN is the use of convolutional layers and kernels.…”

Section: Convolutional Neural Network (Cnn)mentioning

confidence: 99%

“…Realizing the important role of these elements in determining the performance accuracy of phishing detection models, most researchers paid more attention to specifying these parameters, not others such as learning rate, dropout rate, epoch, or batch size. While this problem was avoided in [10], details of optimizing these parameters were not provided in the paper. Similarly, the authors of [24,28,29,32] described the optimization process, but only on certain parameters, for example, the number of convolutional layers, number of kernels, and kernel size.…”

Section: Convolutional Neural Network (Cnn)mentioning

confidence: 99%

“…Similar to DNN and CNN, LSTM can be implemented individually [20,[41][42][43][44][45], incorporated with traditional machine learning techniques [46,47], or combined with other DL algorithms in a hybrid model for an improved performance in detecting malicious websites [10,11,31,33,35,36]. Among the studies of LSTM-based phishing detection models, a majority of them specified the parameter settings for neural network architecture, number of epochs, and learning rate; but ignored the dropout rate and batch size [31,41,42,44,47].…”

Section: Long Short-term Memory (Lstm)mentioning

confidence: 99%

“…For instance, when a dropout rate of 0.2 is applied to a specific layer, 20% of the total number of neurons in that particular layer will be dropped. Dropout strategy is usually used in CNN, LSTM, and GRU architectures to prevent overfitting issues [10,30,34].…”

Section: Hyper-parametersmentioning

confidence: 99%

“…Among them, website phishing detection based on DL algorithms has caught much attention in recent studies. Security strategies based on DL mechanisms have become increasingly popular to deal with evolving phishing attacks [9][10][11]. There are numerous types of DL techniques designed to solve a specific problem or meet a system's particular requirement; each has its advantages and disadvantages [2,12,13].…”

Section: Introductionmentioning

confidence: 99%

See 4 more Smart Citations

Phishing Webpage Classification via Deep Learning-Based Algorithms: An Empirical Study

Selamat

Krejcar

et al. 2021

Applied Sciences

View full text Add to dashboard Cite

Phishing detection with high-performance accuracy and low computational complexity has always been a topic of great interest. New technologies have been developed to improve the phishing detection rate and reduce computational constraints in recent years. However, one solution is insufficient to address all problems caused by attackers in cyberspace. Therefore, the primary objective of this paper is to analyze the performance of various deep learning algorithms in detecting phishing activities. This analysis will help organizations or individuals select and adopt the proper solution according to their technological needs and specific applications’ requirements to fight against phishing attacks. In this regard, an empirical study was conducted using four different deep learning algorithms, including deep neural network (DNN), convolutional neural network (CNN), Long Short-Term Memory (LSTM), and gated recurrent unit (GRU). To analyze the behaviors of these deep learning architectures, extensive experiments were carried out to examine the impact of parameter tuning on the performance accuracy of the deep learning models. In addition, various performance metrics were measured to evaluate the effectiveness and feasibility of DL models in detecting phishing activities. The results obtained from the experiments showed that no single DL algorithm achieved the best measures across all performance metrics. The empirical findings from this paper also manifest several issues and suggest future research directions related to deep learning in the phishing detection domain.

show abstract