In the Fraunhofer-wide project SEC-Learn (Sensor Edge Cloud for Federated Learning), the main aim is to determine which developments are necessary in order that neural networks can be trained directly at the sensor, while at the same time all other sensor nodes benefit from the learned information - so-called federated learning. With local data processing, data protection requirements can be better guaranteed, which should also lead to a higher acceptance by the user. The use of Privacy Enhancing Technologies is intended to create secure data transmission.
The energy consumption, which is necessary for the continuous analysis of the data with AI, is supposed to be reduced by Spiked Neural Networks. These will be implemented energy-efficiently and close to the hardware. For this, methods have to be found which convert the models created by federated learning and make them compatible with the target hardware.