Embodiments are disclosed for reducing unauthorized modifications of object detection systems of motor vehicles. As an example, a method comprises: generating a cryptographic signature for an output of nodes of a fully connected layer of a neural network of an object detection system of a vehicle, the cryptographic signature based in part on a first private key stored in a reply-protected memory block (RPMB), the output at least partially describing a detected object; and, responsive to verifying the cryptographic signature, adjusting vehicle operation based on the detected object. In this way, outputs of the neural network may be verified, so that unauthorized modifications are detected, and system accuracy increases.
