Deep Learning for Phishing Detection

Phishing attacks have caused significant damage to society. With the popularity of two-dimensional codes, a new type of phishing method called two-dimensional code phishing attacks has been created, which targets mobile users. In recent years, with the popularity of mobile devices, the threat of this new type of phishing attack is increasing, and may even exceed the traditional browser-based phishing attacks. In response to this threat, this paper proposes a relative detection method. The legitimacy of the URL in the two-dimensional code is evaluated by using a legal logo embedded in the two-dimensional code. The difficulty of this method is that the logo image is small, and the recognition rate of the small object is reduced by the traditional object recognition method. This paper uses the improved Faster R-CNN for smallscale identification and measures the impact of this method on the FlickrLogos-32 dataset. The experimental results show the effectiveness of the method in logo recognition, which can be used for two-dimensional code phishing attack detection.