Machine Learning-Based Detection of API Security Attacks

The Application Programming Interface provides multiple functionalities in the software development task. Collaboration from third-party developers can also be achieved by using Application Programming Interface. Attackers often target network infrastructure to take advantage of system vulnerabilities for fetchingthe client's sensitive information or any other machine in the network. In some cases, the attackers can use Application Programming Interface for phishing attacks by spoofing like an authorized interface. This research paper presented a comparative analysis of four supervised machine learning techniques for discovering API attacks early so that the client machine should only respond to the authenticated interface discarding the malicious interface. The results achieved for the evaluation of the algorithm Random Forest Classifier, logistic regression techniques, support vector machine, and K-Nearest neighbour are compared with metric accuracy by computing the confusion matrix. The Random Forest classification and logistic regression techniques outperform the other two supervised learning techniques for the traffic dataset for Application Programming Interfaces and the achieved accuracy is close to 98%. The client machine can be trained to detect malicious application programming interfaces by using supervised machine learning techniques like Random Forest Classifier and logistic regression.