No abstract

Federated learning is a novel distributed learning framework, which enables thousands of participants to collaboratively construct a deep learning model. In order to protect confidentiality of the training data, the shared information between server and participants are only limited to model parameters. However, this setting is vulnerable to model poisoning attack, since the participants have permission to modify the model parameters. In this paper, we perform systematic investigation for such threats in federated learning and propose a novel optimization-based model poisoning attack. Different from existing methods, we primarily focus on the effectiveness, persistence and stealth of attacks. Numerical experiments demonstrate that the proposed method can not only achieve high attack success rate, but it is also stealthy enough to bypass two existing defense methods.

No abstract

In this paper, a novel distributed algorithm derived from the event-triggered strategy is proposed for achieving resilient consensus of multi-agent networks (MANs) under deception attacks. These malicious deception attacks are intended to interfere with the communication channel causing periods in time at which the sending information among nodes is modified. In particular, we develop an event-triggered update rule which can mitigate the influence of the attackers and at the same time reduce the computing and communication consumption. Each node chooses the instances to update its state information by checking whether its neighbor set meets a given cardinality-dependent function or not. With specified prerequisite on the coupling weights and the sampling period, the consensus achievement of the MANs is independent of the deception attacks, but strictly depends on the robustness of the interconnection topology. Simulation examples are finally given to illustrate the efficacy of the theoretical results. INDEX TERMS Multi-agent networks, resilient consensus, event-triggered, deception attack.

No abstract