Edge machine learning (Edge ML) offers solutions for deploying ML models directly on resource-constrained edge devices. However, ensuring adversarial robustness remains a challenge. This paper presents an accessible approach for adversarial robust distillation (ARD) based in the limited confines of Google Colab.Our goal is enabling fast yet robust knowledge transfer to student models suited for edge devices. Extensive experiments are conducted distilling from a WideResNet34 teacher to MobileNetV2 student using limited computational resources. The efficacy of ARD is evaluated under settings with only 1 GPU (T4 GPU) and 13GB RAM for up to 6 hours a day.Notably, competitive adversarial robustness is attained using very few gradient attack steps. This improves training efficiency crucial for edge ML. Appropriately balancing hyperparameters also allows robust accuracy over 50% using just 1 attack step. Overall, the presented approach advances the feasibility of performing robust distillation effectively even with accessibility constraints.The democratized and reproducible method on Google Colab serves as a launchpad for those aiming to reap the advantages of edge intelligence. By sharing models protected against adversarial threats, this work propels broader adoption of trustworthy ML at society’s technological edges.