Detecting API Abuse and Authentication Attacks Using Artificial Intelligence in Cloud Applications

Main Article Content

Sai Kiran Arcot Ramesh

Abstract

With cloud applications playing an important role in current businesses, protecting APIs and authentication systems has emerged as the most crucial aspect in preventing abuse and cyberattacks. The paper introduces an innovative hybrid AI-based API abuse and authentication attacks detector in cloud computing. The recommended model is based on the advantages of the random forest (RF) and the Long Short-Term Memory (LSTM) network to identify both the static and the sequential attack patterns. The Random Forest model is very efficient in detecting highvolume anomalies, including rate-limiting attacks and DoS attacks, whereas the LSTM network is very efficient at detecting the time dependencies in the data, being especially well-suited to detect brute-force attacks and credential stuffing attacks. The hybrid model proved to be a better model than the single models with 96.5% accuracy, 93% precision and 98% recall and AUC, 0.97. The model was shown to be superior to the traditional rule-based systems, pointing to the benefits of ensemble learning and deep learning to address the complexity and continually changing nature of current cyber threats. The model has challenges observed in terms of computational complexity as well as real-world datasets, despite performance which is promising. The next step in work will be to increase scalability, solve false positive and false negative rates, and add other types of attacks to a more complete security solution. This study contributes to a powerful, AI-based method of real-time security of cloud applications, and it would offer a strong opportunity to enhance the detection and prevention of API abuse and authentication attacks.

Article Details

Section
Articles