Nowadays, applications of Convolutional Neural Networks are common in sensible contexts. These specific neural networks are powerful and adaptable to many computer vision tasks, but they are also vulnerable to adversarial attacks, therefore, also imperceptible perturbations added to legit input images can cause a network to fail. Some techniques have been proposed to mitigate this risk: the challenge is in the black box nature of Deep Neural Networks. In this presentation, the speaker will pick up one of the most advanced mitigation techniques and show how Explainable AI can help validating it.

• Convolutional Neural Networks (CNN) overview
• Adversarial attacks to CNNs
• Explainable AI (XAI) 101
• Validation of mitigation defense through XAI

5G has been fuelling EU headlines for the past years. As a result, EU commission published Risk Assessment Report and ENISA wrote a 5G Threat Landscape. Both documents are highlighting the importance of collaboration and sharing information between the different actors.Building the necessary trust goes beyond the words, Huawei will present its view on how to collaborate with ecosystem stakeholders to raise E2E security.

• Example of a risk-based approach for a vertical sector
• Setting up collaboration within an ecosystem and establishing trustworthiness
• Role of standards and regulations