Hands-On Artificial Intelligence for Cybersecurity

Neural network vulnerabilities

Despite the fact, as we have seen previously, that NNs have acquired particular relevance in recent times (as we have seen previously), due to their significant potential when it comes to resolving more complex problems that are usually the prerogative of human cognitive abilities, such as facial recognition and speech recognition, NNs, especially DNNs, suffer from a number of rather important vulnerabilities, which can be exploited through the use of GANs. This implies the possibility, for example, of deceiving biometric authentication procedures based on facial recognition or other biometric evidence made possible by the artificial creation of adversarial examples.

Apparently harmless devices such as 3D medical ...