BLIA: Detect model memorization in binary classification model through passive Label Inference attack