This paper introduces a practical black-box attack against machine learning models, specifically deep neural networks (DNNs), without requiring knowledge of the model's internal structure or training data. The attack leverages only the ability to observe labels assigned by the DNN to chosen inputs, akin to a cryptographic oracle. The strategy involves training a local substitute model using synthetically generated inputs and labels provided by the target DNN. Adversarial examples are then crafted using this substitute model, which are misclassified by the target DNN due to similar decision boundaries. The attack is validated against DNNs hosted by MetaMind, Amazon, and Google, achieving high misclassification rates. The attack is also shown to evade defenses previously designed to counter adversarial examples. The method is applicable to various ML techniques, including logistic regression, and demonstrates the vulnerability of real-world systems to adversarial attacks. The paper highlights the importance of robust security measures against such threats.This paper introduces a practical black-box attack against machine learning models, specifically deep neural networks (DNNs), without requiring knowledge of the model's internal structure or training data. The attack leverages only the ability to observe labels assigned by the DNN to chosen inputs, akin to a cryptographic oracle. The strategy involves training a local substitute model using synthetically generated inputs and labels provided by the target DNN. Adversarial examples are then crafted using this substitute model, which are misclassified by the target DNN due to similar decision boundaries. The attack is validated against DNNs hosted by MetaMind, Amazon, and Google, achieving high misclassification rates. The attack is also shown to evade defenses previously designed to counter adversarial examples. The method is applicable to various ML techniques, including logistic regression, and demonstrates the vulnerability of real-world systems to adversarial attacks. The paper highlights the importance of robust security measures against such threats.