Federated Learning (FL) is nowadays one of the most promising paradigms for privacy-preserving distributed learning. Without revealing its local private data to outsiders, a client in FL systems collaborates to build a global Deep Neural Network (DNN) by submitting its local model parameter update to a central server for iterative aggregation. With secure multi-party computation protocols, the submitted update ofany clientis also by design invisible to the server. Seemingly, this standard design is a win-win for client privacy and service provider utility. Ironically, any attacker may also use manipulated or impersonated client to submit almost any attack payloads under the umbrella of the FL protocol itself. In this work, we craft a practical backdoor attack on FL systems that is proved to be simultaneously effective and stealthy on diverse use cases of FL systems and leading commercial FL platforms in the real world. Basically, we first identify a small number of redundant neurons which tend to be rarely or slightly updated in the model, and then inject backdoor into these redundant neurons instead of the whole model. In this way, our backdoor attack can achieve a high attack success rate with a minor impact on the accuracy of the original task. As countermeasures, we further consider several common technical choices including robust aggregation mechanisms, differential privacy mechanism,s and network pruning. However, none of the defenses show desirable defense capability against our backdoor attack. Our results strongly highlight the vulnerability of existing FL systems against backdoor attacks and the urgent need to develop more effective defense mechanisms.

Following the principle of to set one's own spear against one's own shield, we study how to design adversarial completely automated public turing test to tell computers and humans apart (CAPTCHA) in this article. We first identify the similarity and difference between adversarial CAPTCHA generation and existing hot adversarial example (image) generation research. Then, we propose a framework for text-based and image-based adversarial CAPTCHA generation on top of state-of-the-art adversarial image generation techniques. Finally, we design and implement an adversarial CAPTCHA generation and evaluation system, called aCAPTCHA, which integrates 12 image preprocessing techniques, nine CAPTCHA attacks, four baseline adversarial CAPTCHA generation methods, and eight new adversarial CAPTCHA generation methods. To examine the performance of aCAPTCHA, extensive security and usability evaluations are conducted. The results demonstrate that the generated adversarial CAPTCHAs can significantly improve the security of normal CAPTCHAs while maintaining similar usability. To facilitate the CAPTCHA security research, we also open source the aCAPTCHA system, including the source code, trained models, datasets, and the usability evaluation interfaces.

The development of deep learning techniques has significantly increased the ability of computers to recognize CAPTCHA (Completely Automated Public Turing test to tell Computers and Humans Apart), thus breaking or mitigating the security of existing captcha schemes. To protect against these attacks, recent works have been proposed to leverage adversarial machine learning to perturb captcha pictures. However, they either require the prior knowledge of captcha solving models or lack adaptivity to the evolving behaviors of attackers. Most importantly, none of them has been deployed in practical applications, and their practical applicability and effectiveness are unknown.