2020 American Control Conference (ACC) 2020
DOI: 10.23919/acc45564.2020.9147690
|View full text |Cite
|
Sign up to set email alerts
|

Parameter Privacy versus Control Performance: Fisher Information Regularized Control

Help me understand this report

Search citation statements

Order By: Relevance

Paper Sections

Select...
2
1
1
1

Citation Types

0
10
0

Year Published

2020
2020
2024
2024

Publication Types

Select...
4
4

Relationship

3
5

Authors

Journals

citations
Cited by 12 publications
(10 citation statements)
references
References 13 publications
0
10
0
Order By: Relevance
“…Remark 4: We have introduced the expectation of a lower bound of Λ−1 because the computation of the inverse of Λ, in general, requires a large number of computation resources, and it cannot be computed in advance of the control system's operation. A similar approach can be found in [40]- [43], and the studies employed the inverse of a trace of the Fisher information matrix as a lower bound of the precision of general unbiased estimator for dynamical systems. Unfortunately, the approach is not specialized in our attack scenario, i.e., it would give a loose lower bound of tr( Λ−1 ), and the lower bound cannot be computed without the system's operating data.…”
Section: A Sample Identifying-complexity Curvementioning
confidence: 99%
“…Remark 4: We have introduced the expectation of a lower bound of Λ−1 because the computation of the inverse of Λ, in general, requires a large number of computation resources, and it cannot be computed in advance of the control system's operation. A similar approach can be found in [40]- [43], and the studies employed the inverse of a trace of the Fisher information matrix as a lower bound of the precision of general unbiased estimator for dynamical systems. Unfortunately, the approach is not specialized in our attack scenario, i.e., it would give a loose lower bound of tr( Λ−1 ), and the lower bound cannot be computed without the system's operating data.…”
Section: A Sample Identifying-complexity Curvementioning
confidence: 99%
“…The statistical parameter privacy problem has been studied in [21] and [22]. Bassi et al in [21] studied the statistical parameter privacy of an independent and identically distributed (i.i.d.…”
Section: B Related Workmentioning
confidence: 99%
“…They characterized the leakage level of private information under the Bayes statistical risk as the privacy measure. The authors in [22] studied the optimal design of controller and privacy filter for a linear Gaussian plant. The system dynamics should be kept private from an adversary interested in inferring the system dynamics based on the state measurements and control inputs.…”
Section: B Related Workmentioning
confidence: 99%
See 1 more Smart Citation
“…• On a more technical note, our SDP characterization of the distribution optimizing the trace of inverse Fisher information may also be of independent interest, as this quantity also plays a key role in certain privacypreserving mechanisms, e.g. in [16], [17], [18]. By contrast, previous results were only able to characterize this distribution for scalar random variables, [10], or considered relaxed problems [16].…”
Section: Introductionmentioning
confidence: 99%