Federation defense method based on AIoT-oriented security

The invention discloses a federated defense method for AIoT (Ad Hoc Internet of Things) security. The method is characterized by comprising the following steps of: each edge end node collects countermeasure attack samples encountered locally, and uses the normal samples and the countermeasure samples to carry out countermeasure training independently and sends the gradient information updated by each round of adversarial training to the cloud, and the cloud performs parameter updating once by using the aggregated gradient to complete model parameter synchronization in the whole distributed defense architecture. Compared with the prior art, the method has certain defense universality; when the neural network on any AIoT device is attacked by an adversarial sample, defensive repair is carried out on the attacked AIoT equipment on the premise of ensuring the privacy security of the AIoT terminal, so that the problems that the specific training data of the AIoT environment is unbalanced, the potential attack coun