A Character Level Word Encoding Deep Learning Model for Combating Cyber Threats in Phishing URL Detection
Keywords:
Cyber Security, Cyber-crimes, Data Sanitization, Tokenization, Machine Learning, Deep LearningAbstract
A cyber threat is generally a malicious activity which tend to damage or steal data or in general something which disrupts the digital life. Security errors, DoS attacks, malware, and data theft are some of these dangers. A kind of cyber threat known as "phishing" occurs when attackers impersonate a legitimate URL or website to collect user information. Out of the total cyber-crimes reported in last quarter around the globe, 21% falls in the phishing category. Phishing is often performed as a social engineering method and the conventional detection techniques were largely relied on the manual reports. Recently, techniques for machine learning have been used to identify phishing. Owing to the recent advancement in the deep learning methods, many possibilities have also been discussed on using the same for achieving better performance. To identify malicious URLs, this research presents a lightweight deep-learning model that can function quite efficient enough even in energy-saving systems. The experimental results of the proposed model showed substantial improvements considering the parameters of comparison to other cutting-edge methods. There was an enhancement in the rate of accurately detecting true positives. Furthermore, the experiments confirmed that the proposed approach operates quite efficiently even when phishing detection systems are in energy-saving modes