Search

The rapidly expanding number of Internet of Things (IoT) devices is generating huge quantities of data, but the data privacy and security exposure in IoT devices, especially in the automatic driving system. Federated learning (FL) is a paradigm that addresses data privacy, security, access rights, and access to heterogeneous message issues by integrating a global model based on distributed nodes. However, data poisoning attacks on FL can undermine the benefits, destroying the global model's availability and disrupting model training. To avoid the above issues, we build up a hierarchical defense data poisoning (HDDP) system framework to defend against data poisoning attacks in FL, which monitors each local model of individual nodes via abnormal detection to remove the malicious clients. Whether the poisoning defense server has a trusted test dataset, we design the \underline{l}ocal \underline{m}odel \underline{t}est \underline{v}oting (LMTV) and \underline{k}ullback-\underline{l}eibler divergence \underline{a}nomaly parameters \underline{d}etection (KLAD) algorithms to defend against label-flipping poisoning attacks. Specifically, the trusted test dataset is utilized to obtain the evaluation results for each classification to recognize the malicious clients in LMTV. More importantly, we adopt the kullback leibler divergence to measure the similarity between local models without the trusted test dataset in KLAD. Finally, through extensive evaluations and against the various label-flipping poisoning attacks, LMTV and KLAD algorithms could achieve the $100\%$ and $40\%$ to $85\%$ successful defense ratios under different detection situations.

Starting from the local structures to study hierarchical trees is a common research method. However, the cumbersome analysis and description make the naive method challenging to adapt to the increasingly complex hierarchical tree problems. To improve the efficiency of hierarchical tree research, we propose an embeddable matrix representation for hierarchical trees, called Generation Matrix. It can transform the abstract hierarchical tree into a concrete matrix representation and then take the hierarchical tree as a whole to study, which dramatically reduces the complexity of research. Mathematical analysis shows that Generation Matrix can simulate various recursive algorithms without accessing local structures and provides a variety of interpretable matrix operations to support the research of hierarchical trees. Applying Generation Matrix to differential privacy hierarchical tree release, we propose a Generation Matrix-based optimally consistent release algorithm (GMC). It provides an exceptionally concise process description so that we can describe its core steps as a simple matrix expression rather than multiple complicated recursive processes like existing algorithms. Our experiments show that GMC takes only a few seconds to complete a release for large-scale datasets with more than 10 million nodes. The calculation efficiency is increased by up to 100 times compared with the state-of-the-art schemes., Comment: 25 pages, 10 figures

Adaptive control technique is adopted to synchronize two identical non-autonomous systems with unknown parameters in finite time. A virtual unknown parameter is introduced in order to avoid the unknown parameters from appearing in the controllers and parameters update laws. The Duffing equation and a gyrostat system are chosen as the numerical examples to show the validity of the present method., Comment: 6 pages, 4 figures.Submitted to The 8th IEEE International Conference on Control & Automation

In this paper we investigate a master-slave synchronization scheme of two n-dimensional non-autonomous chaotic systems coupled by sinusoidal state error feedback control, where parameter mismatch exists between the external harmonic excitation of master system and that of slave one. A concept of synchronization with error bound is introduced due to parameter mismatch, and then the bounds of synchronization error are estimated analytically. Some synchronization criteria are firstly obtained in the form of matrix inequalities by the Lyapunov direct method, and then simplified into some algebraic inequalities by the Gerschgorin disc theorem. The relationship between the estimated synchronization error bound and system parameters reveals that the synchronization error can be controlled as small as possible by increasing the coupling strength or decreasing the magnitude of mismatch. A three-dimensional gyrostat system is chosen as an example to verify the effectiveness of these criteria, and the estimated synchronization error bounds are compared with the numerical error bounds. Both the theoretical and numerical results show that the present sinusoidal state error feedback control is effective for the synchronization. Numerical examples verify that the present control is robust against amplitude or phase mismatch., Comment: 30 pages, 10 figures, accepted for publication in International Journal of Modern Physics B