An Adversary-Resilient Doubly Compressed Diffusion LMS Algorithm for Distributed Estimation

Zayyani, Hadi; Oruji, Fatemeh; Fijalkow, Inbar

doi:10.1007/s00034-022-02072-w

Cited by 12 publications

(3 citation statements)

References 33 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…In this section, the simulation results are presented. The network used in the simulation has N = 16 agents and is similar to that used in [11] and is depicted in Fig. 1.…”

Section: Simulation Resultsmentioning

confidence: 99%

Privacy-Preserving Distributed Estimation via Laplacian and Gaussian Differential-Privacy Noise

Zayyani

2024

Preprint

Self Cite

View full text Add to dashboard Cite

In this paper, a private distributed estimation algorithm is proposed. In this algorithm, a differential-privacy noise is added to the intermediate estimation to be exchanged among nodes. Two types of differential noise is regarded in the paper which are Gaussian and Laplacian. Also, in each case, two approaches are used to recover the true intermediate estimations. In the first approach, we estimate the true intermediate estimation and in the second approach, we estimate the noise vector and then subtract it from the noise intermediate estimation. We show that both approaches lead to the same formula for denoised intermediate estimation. Simulation experiments corroborate the effectiveness of the proposed algorithm when variance of added privacy noise is high and the privacy is guaranteed with high confidence.

show abstract

“…In this section, the simulation results are presented. The network used in the simulation has N = 16 agents and is similar to that used in [11] and is depicted in Fig. 1.…”

Section: Simulation Resultsmentioning

confidence: 99%

Privacy-Preserving Distributed Estimation via Laplacian and Gaussian Differential-Privacy Noise

Zayyani

2024

Preprint

Self Cite

View full text Add to dashboard Cite

show abstract

“…The network has N = 16 sensors and the topology is the same as was used in [32] and depicted in Fig. 5.…”

Section: Simulation Resultsmentioning

confidence: 99%

Hidden Markov Model Decorrelated Diffusion Leaky LMS With Optimized Leaky Factor

Zayyani,

Salman,

Korki

et al. 2024

IEEE Access

Self Cite

View full text Add to dashboard Cite

In this paper, two basic novel modifications of Diffusion Leaky Least Mean Square (DL-LMS) algorithm are proposed. First, the optimized leaky factor is obtained in closed form using the minimum disturbance principle, thanks to quadratic form of disturbance. Then, a low complexity decorrelated version of DL-LMS is presented based on the statistical Hidden Markov Modeling (HMM) of time correlations of input signals. The decorrelation is performed using a simple one-tap filter. Simulation results demonstrate the effectiveness of both modifications in comparison to other methods. INDEX TERMS Distributed estimation, leaky LMS, hidden Markov model, diffusion.

show abstract

“…One way is setting the elastic fusion weight to lower the impact of malicious nodes, and thus the estimator is robust to attacks [2][3][4][5]. Another approach is to add attack detection step before data fusion into existing distributed estimation algorithms [6][7][8][9][10][11]. For instance, reference [6] proposes a secure diffusion least mean square (S-dLMS) algorithm, in which each node detects adversarial neighbours by sorting the variables received.…”

mentioning

confidence: 99%

Temporal convex combination‐based secure distributed estimation against cyberattacks and noisy input

Zhang,

Jia,

Peng

et al. 2024

Electronics Letters

View full text Add to dashboard Cite

This paper proposes a new secure distributed estimation for cyber‐physical systems against adversarial attacks with noisy input. To mitigate the effect of attacks, a novel distributed attack detection based on a reliable reference estimation obtained by temporal convex combination is proposed. Furthermore, for more effective and robust performance, an adaptation rule to adjust convex combination weights is presented, in which the generalized correntropy method with nonlinear loss function and stochastic gradient descent are utilized. Besides, to eliminate input noise, a bias‐compensation method in local adaptation of the secure distributed estimation is proposed. Simulations show superior dynamic and real‐time adaptability of the proposed algorithm under complex attacking scenarios.

show abstract

An Adversary-Resilient Doubly Compressed Diffusion LMS Algorithm for Distributed Estimation

Cited by 12 publications

References 33 publications

Privacy-Preserving Distributed Estimation via Laplacian and Gaussian Differential-Privacy Noise

Privacy-Preserving Distributed Estimation via Laplacian and Gaussian Differential-Privacy Noise

Hidden Markov Model Decorrelated Diffusion Leaky LMS With Optimized Leaky Factor

Temporal convex combination‐based secure distributed estimation against cyberattacks and noisy input

Contact Info

Product

Resources

About