Logo Image Based Approach for Phishing Detection

Phishing is a cyber attack which involves a fake website mimicking the some real legitimate website. The website makes the user believe the website being authentic and thus online user provides their sensitive information like password, PIN, Social Security Number, and Credit Card Information etc. Due to involvement of such high sensitivity information, these websites are a huge threat to online users and detection and blocking of such website become crucial. In this thesis, we propose a new phishing detection method to protect the internet users from such attacks. In particular, given a website, our proposed method will be able to detect between a phishing website and a legitimate website just by the screenshot of the logo image of it. Due to the usage of screenshot for extracting the logo, any hidden logo will not be able to spoof the algorithm into considering the website as phishing as happened in existing methods. In first study focus was on dataset gathering and then the logo image is extracted. This logo image is uploaded to Google image search engine using automated script which returns the URLs associated with that image. Since the relationship between logo and domain name is exclusive it is reasonable to treat the logo image as identity of original URL. Hence the phishing website will not have the same relation to the logo image as such and will not get returned as URL by Google when search for that logo image. Further, Alexa page rank is also used to strengthen the detection accuracy.