Federated learning–based IoT malware detection using the IoT-23 dataset, evaluated under adversarial settings including label flipping, gradient manipulation, sign flipping, and single-client backdoor attacks, with time-aware preprocessing and calibration analysis.