Detection of Fake URLs Using Deep LSTM Architecture over Social Media

In the era of Online Social Networks (OSNs), efficient communication is crucial for disseminating information to audiences swiftly. However, this convenience comes with challenges as Clickbait and malicious activities pose significant risks to users and such cybercrimes result in substantial financial losses annually. Addressing this issue is paramount to safeguard users and maintain the integrity of OSNs. This study proposes a two-level Uniform Resource Locator(URL) detection technique to combat these threats. Firstly, a human-annotated database is created from various phishing websites, facilitating the differentiation between authentic and fraudulent URLs. Then, a deep Long-Short Term Memory (LSTM) neural network, leveraging a 100-dimensional Glove word embedding at the word level, extracts feature vectors for URLs which enables accurate predictions regarding URL authenticity. Additionally, a ground dataset is established, categorizing URLs as genuine or fake based on trustworthy websites.