A system for detecting phishing events is provided. A data receiver is configured to receive datasets representative of web traffic associated with access to or on-going usage of an application hosted on a server of a production environment by a user. A machine learning engine is configured to generate a score based at least on the datasets representative of the web traffic indicative of whether the user is a malicious user or a non-malicious user. A routing modification engine is configured to route downstream web traffic associated with access to or on-going usage of the application by the user if the score is greater than a threshold to a server of a sandbox environment that is configured to emulate a graphic user interface of the production environment.