PDRCNN: Precise Phishing Detection with Recurrent Convolutional Neural Networks

Table 1

Comparison of PDRCNN with related works.

Type	Work	Search engine dependence	Third-party dependence	Language dependence	Number of experimental samples
Blacklist/whitelist-based	Google Safe Browsing API [2]	No	Yes	No	–/–
	AIWL [3]	No	No	No	16/18
Visual similarity-based	Doom Tree similarity [4]	No	No	No	8/320
	BaitAlarm [5]	Yes	Yes	No	0/300
	LinkGuard [6]	No	Yes	No	0/8
	Phishdentity [7]	Yes	Yes	No	5000/5000
Heuristic-based	CANTINA [8]	Yes	Yes	Yes (English)	100/100
	PhishNet [9]	No	Yes	No	0/6000
	Finite state machine [10]	Yes	No	No	99/25
	New approach [11]	Yes	Yes	Yes (English)	1200/3374
	PhishShield [12]	No	Yes	No	250/1600
	PDA [13]	No	Yes	No	405/1120
Machine learning-based	Fuzzy logic [14]	Yes	Yes	No	0/606
	CANTINA+ [15]	Yes	Yes	Yes (English)	4883/8118
	Page classification [16]	Yes	No	No	200/325
	AC [17]	No	Yes	No	450/2500
	MCAC [18]	No	Yes	No	1350 (All)
	SMO [19]	No	Yes	Yes (Chinese)	1462/1416
	Phish detector [20]	Yes	Yes	No	1271/3066
	Know thy domain name [21]	No	Yes	No	2000/4013
	Metaheuristic algorithm [22]	Yes	Yes	No	8599/2456
	HEFS [23]	No	No	No	5000/5000
Deep learning-based	Classifying phishing URLs using RNN [24]	No	No	No	1000000/1000000
	Stacked autoencoder [26]	Yes	Yes	No	20000/17000
	Phishing detection with LSTM [25]	No	Yes	No	2000/2000