AI Powered Phishing Link Identifier for Social Media DMs | IJCT Volume 12 – Issue 6 | IJCT-V12I6P31
International Journal of Computer Techniques
ISSN 2394-2231
Author
Nayana H S, Harshitha R, Namitha Biswal, Mahesh Gowda N S, Dr Guruprasad Y K
Abstract
In today’s digital landscape, social media platforms have become a medium for cyber attackers to perform malicious activities. Attackers use Direct Messages to trick users into revealing their identities by clicking on the malicious links. As modern phishing techniques are evolving, conventional filters such as fixed rule-based filters, URL static blacklists have become ineffective.
The project AI-powered phishing link identifier for social media DMs aims to develop a system that detects and warns phishing URLs shared through social media DMs using XG Boost model. It evaluates features including domain length, special characters, entropy, HTTPS presence to analyze phishing links. The XG Boost model has 98% of accuracy, performs better than other classifiers such as Random Forests, SVM, Logistic Regression.
This detection shows how AI can identify and reduce the malicious activities like phishing in social media platforms providing real time solutions effectively.
Keywords
^KEYWORDS^
Conclusion
Phishing is still one of the biggest and most changing cybersecurity problems. It takes advantage of how people think and the trust they have in digital messages. As phishing attacks move from regular email to direct messages on social media, there is a bigger need for smart, quick, and private ways to spot these threats. This research introduced an AI-based system for finding phishing links using the XG Boost machine learning model.
It was made for social media settings. The system looks at the structure, words, and patterns in URLs to tell if a link is bad or good. It doesn’t need to check the message content or use outside information. This makes it fast, efficient, and safe for use on messaging apps. Tests showed the XG Boost system was very good at identifying phishing links, with an accuracy of 98.1%. It did better than older machine learning models and was almost as good as deep learning methods, but used less computer power. The model is also easy to use, can work with different kinds of data, and handles new phishing tricks well.
Besides being strong in performance, this system shows a forward-thinking approach to cybersecurity—focusing on stopping attacks before they happen. It can be added to social media sites, work networks, web browsers, and mobile apps to stop phishing in real time. This study helps fix important issues like not enough research on direct messages, reliance on outside data, and the need for fast detection. It gives a real and scalable solution for modern phishing problems.
In the end, this system shows that AI models like XG Boost can be a strong part of future phishing defenses. They offer accurate, clear, and quick protection across online platforms. Future work will improve it with deep hybrid learning, explainable AI, and shared model updates. This could lead to smarter, more automatic, and privacy-focused ways to stop phishing in a world where social media is everywhere.
References
[1] A. Dalsaniya, “AI-Based Phishing Detection Systems: Real-Time Email and URL Classification,” The International Journal of Engineering Research (TIJER), vol. 12, no. 5, pp. 45–52, 2023.
[2] O. I. Enitan, “An AI-Powered Approach to Real-Time Phishing Detection,” International Journal of Modern Computer and Information Systems (IJMCIS), vol. 8, no. 2, pp. 112–120, 2023.
[3] O. A. Lamina, and P. Broklyn, “AI-Powered Phishing Detection and Prevention,” Path of Science, vol. 10, no. 6, pp. 45–54, 2024.
[4] F. Basit, S. Zafar, and M. J. Khan, “Intelligent Phishing Detection via Ensemble Learning,” IEEE Access, vol. 9, pp. 56320–56332, 2021.
[5] B. Gupta, A. Tewari, and A. Jain, “Fighting Phishing Attacks: A Survey of Existing Techniques,” Computers & Security, vol. 103, pp. 1–25, 2022.
[6] X. Zhang, L. Li, and J. Liu, “Hybrid Machine Learning Models for Phishing Website Detection,” Information Sciences, vol. 628, pp. 475–490, 2023.
[7] D. Parmar, S. Shah, and K. Patel, “Advanced AI in Social Media Threat Detection,” Elsevier Journal of Information Security, vol. 17, no. 4, pp. 211–223, 2023.
[8] F. Soomro, Z. Hussain, and M. Karim, “Explainable AI for Cyber Threat Detection,” IEEE Transactions on Information Forensics and Security, vol. 19, pp. 1140–1151, 2024.
[9] K. Mahmoud and H. Mahfouz, “Image Recognition Techniques for Phishing Detection,” International Journal of Computer Science (IJCS), vol. 15, no. 2, pp. 118–127, 2020.
[10] R. Chiew, K. Yong, and C. Tan, “PhishFinder: Machine Learning-Based Phishing URL Detection,” Expert Systems with Applications, vol. 107, pp. 11–21, 2022.
[11] A. Jain and V. Gupta, “Comparative Study of URL-Based Machine Learning Models for Phishing Detection,” Procedia Computer Science, vol. 218, pp. 1500–1508, 2023.
[12] H. Lee, T. Nguyen, and J. Kim, “Real-Time Phishing URL Detection Based on Machine Learning,” Sensors, vol. 22, no. 7, pp. 2556–2565, 2022.
[13] A. Aljabri, N. Alghamdi, and S. Alotaibi, “AI-Powered Cyber Threat Prediction Using XGBoost and Random Forest,” IEEE Access, vol. 10, pp. 120450–120463, 2022.
[14] Y. Yang, Z. Zhang, and P. Liu, “Phishing Detection Using Explainable Gradient Boosting Models,” Journal of Network Security, vol. 18, no. 2, pp. 75–86, 2023.
[15] M. Abbas, S. Mehmood, and U. Iqbal, “Social Engineering and AI-Based Defenses,” Springer Advances in Cybersecurity, vol. 12, no. 1, pp. 91–104, 2021.
[16] S. Aghatise and T. Oke, “AI-Driven Defense Mechanisms for Social Media Phishing,” IEEE International Conference on Cyber Intelligence, pp. 204–210, 2024.
How to Cite This Paper
Nayana H S, Harshitha R, Namitha Biswal, Mahesh Gowda N S, Dr Guruprasad Y K (2025). AI Powered Phishing Link Identifier for Social Media DMs. International Journal of Computer Techniques, 12(6). ISSN: 2394-2231.
