Abstract: The growth of the Internet of Things has been aided by the advancement of artificial intelligence technologies (IoT). However, when using the internet, this promising cyber technology may confront major security issues. A malicious website can impersonate a legitimate website in order to steal users' personal information. As a result, it's critical to use tools like machine learning algorithms to detect fraudulent websites, as these algorithms can make it easier to spot aberrant information concealed in large amounts of data. As a result, many feature engineering activities must be performed from memory, as excellent features substantially improve a powerful machine learning model. We present an unsupervised learning method for learning URL embedding in this work. We also investigate certain important characteristics of a domain embedding model in order to achieve a positive influence on domain features.Keywords—URL embedding model, machine learning, malicious websites, feature engineering.