Abstract:
The detection and mitigation of injection attacks have become crucial due to growing reliance on computer networks and the possibility of cyberattacks in the Internet of ...Show MoreMetadata
Abstract:
The detection and mitigation of injection attacks have become crucial due to growing reliance on computer networks and the possibility of cyberattacks in the Internet of Things(IoT) environment. To improve the security of software programs and systems, this study investigates the use of machine learning(ML) techniques for identifying signs of injection attacks. Through extensive testing and evaluation, different ML methods are trained and evaluated on datasets comprising both valid inputs and injection attacks. Accuracy, recall, precision, and F1-score are performance metrics used to measure how well the models can distinguish between legitimate and malicious inputs. The findings show that the ML algorithms used can effectively identify injection attacks, and the high accuracy rates attained support their potential as effective defence strategies. This study sets the path for future improvements and optimisations in cybersecurity tactics while also advancing our understanding of the viability and effectiveness of ML-based approaches in preventing injection attacks. The incorporation of such ML-based detection devices offers an anticipatory and adaptive solution to reduce the growing threat of injection assaults as digital systems continue to develop. The paper uses the SQL injection attack detection dataset to forecast injection attacks that make use of different ML methods, namely LR, RF, SVM, and NB. The best attack detection method has been determined using performance metrics such as accuracy, sensitivity, precision, and F1-score. The LR, RF and SVM models, with accuracy ratings of 94.40%, 91.19%, and 81.78%, respectively, are followed by the NB model, which has the highest accuracy rating of 97.73 %.
Date of Conference: 24-26 November 2023
Date Added to IEEE Xplore: 22 February 2024
ISBN Information: