Web push notifications are used to induce users to phishing websites. To prevent this induction, keyword-based detection has been proposed. However, the words used in phishing web push notifications differ from those used in phishing websites. Thus, generating a keyword list for detecting phishing web push notifications is challenging. In this paper, we propose a method for detecting phishing-targeted web push notifications through image similarity analysis. Phishing-targeted web push notifications mostly contain images in their data. We utilized this image data to detect phishing-targeted web push notifications. In our prototype, we used mitmproxy to capture image data and analyzed image similarity using structural image similarity. Experimental results showed that our prototype can detect phishing-targeted web push notifications with high accuracy.