Abnormal Client Behavior Detection in Federated Learning
This addresses the issue of minimizing adverse impacts from anomalous clients in federated learning, which is critical for maintaining system integrity, but it appears incremental as it builds on existing detection paradigms.
The paper tackled the problem of detecting abnormal client behaviors in federated learning systems, such as malicious attackers or malfunctioning clients, by proposing a server-side detection method using low-dimensional surrogates of model weight vectors, and it showed that this approach significantly outperforms conventional defense-based methods in experiments on the FEMNIST dataset.
In federated learning systems, clients are autonomous in that their behaviors are not fully governed by the server. Consequently, a client may intentionally or unintentionally deviate from the prescribed course of federated model training, resulting in abnormal behaviors, such as turning into a malicious attacker or a malfunctioning client. Timely detecting those anomalous clients is therefore critical to minimize their adverse impacts. In this work, we propose to detect anomalous clients at the server side. In particular, we generate low-dimensional surrogates of model weight vectors and use them to perform anomaly detection. We evaluate our solution through experiments on image classification model training over the FEMNIST dataset. Experimental results show that the proposed detection-based approach significantly outperforms the conventional defense-based methods.