Machine Learning Approaches for Threat Detection in Cloud Computing
Sai Vamsi Chennupati, Somnath Banerjee
Abstract
This paper provides an in-depth discussion of ML (machine learning) algorithms to identify threats in cloud computing infrastructure, with a specific focus on intelligent detection of anomalies, intrusions, and malicious behaviors in dynamic cloud networking conditions. The suggested architecture applies a hybrid architecture with Convolutional Neural Networks (CNN) to extract features and LSTM (Long Short-Term Memory) networks to evaluate beginning and ending temporal behavior. A database containing <tex xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">$\mathbf{5 0, 0 0 0}$</tex> samples of network traffic (normal operation and threatinduced operation) was referenced in the form of a cloud database. Preprocessing methods such as normalization and feature selection through Recursive FIre Elimination (RFE) method served data preprocessing. Experimental analysis revealed that the hybrid CNN-LSTM model had an Accuracy of 98.3 % and a Precision of 97.6%, a Recall of 98.1, and a False Alarm rate (FAR) of 1.2, although compared with the traditional algorithms like SVM and random forest, the hybrid model highly surpassed them in accuracy, precision, recall, and FAR. The proposed plan is quite adaptable to evolving threat profiles and inter-cross-live relations between virtual machine generations. Overall, the results prove the model is effective in deploying cloud services that are resilient to and secure against advanced attack-based bases.