Data Poisoning Attacks on Federated Using Machine Learning

Data poisoning attacks are a type of adversarial attack that aims to corrupt the training data used to build machine learning models. In this study, we investigate the effectiveness of data poisoning attacks on Three popular machine learning algorithms: SVM, PCA and Naïve Bayes, and Decision We propose a novel data poisoning attack that selectively manipulates training data to induce Miss Classification. Our attack strategy involves injecting a small number of Malicious examples that are designed to bias the decision boundaries of the classifiers towards a specific class. Our experimental results demonstrate that our attack strategy is effective and can significantly degrade the performance of the targeted classifiers. Specifically, our attack achieves a success rate of up to 90% on SVM, PCA and Naïve Bayes, and up to 70% on Decision Tree. Furthermore, we show that our attack is robust to various defenses, including outlier removal and regularization. Our findings highlight the vulnerability of machine learning models to data poisoning attacks and emphasize the need for developing robust and secure machine learning algorithms