QuEry Based Attack
Query-based attacks exploit the vulnerabilities of machine learning models by iteratively querying the model with carefully crafted inputs to infer information or manipulate its behavior, often without direct access to its internal workings. Current research focuses on developing more efficient and effective attack strategies, particularly for large language models and image classifiers, employing optimization algorithms like gradient descent and evolutionary methods, and exploring defenses such as stateful monitoring and randomized input transformations. Understanding and mitigating these attacks is crucial for ensuring the security and reliability of deployed machine learning systems across various applications, from malware detection to protecting user privacy in conversational AI.