Differentially Private Distributed Convex Optimization via Functional Perturbation

Nozari, Erfan; Tallapragada, Pavankumar; Cortés, Jorge

doi:10.1109/tcns.2016.2614100

Cited by 138 publications

(106 citation statements)

References 26 publications

Supporting

Mentioning

103

Contrasting

Order By: Relevance

“…To secure the data and operate on ciphertext space, a homomorphic encryption method is discussed in [20]. Privacy preservation in Machine Learning (ML) is addressed using a differential privacy paradigm, which deals with adding a statistically-designed noise to the exchanged functions or states to protect the sensitive data [21]. Alternatively, in [22], a cryptographic image classification algorithm is proposed on a multi-layer extreme learning system that is capable of specifically classifying encrypted images without decryption.…”

Section: Related Workmentioning

confidence: 99%

Energy-Efficient End-to-End Security for Software-Defined Vehicular Networks

Raja

Anbalagan

Vijayaraghavan

et al. 2021

IEEE Trans. Ind. Inf.

View full text Add to dashboard Cite

One of the most promising application areas of the Industrial Internet of Things (IIoT) is Vehicular Ad hoc NETworks (VANETs). VANETs are largely used by Intelligent Transportation Systems (ITS) to provide smart and safe road transport. To reduce the network burden, Software Defined Networks (SDNs) acts as a remote controller. Motivated by the need for greener IIoT solutions, this paper proposes an energy-efficient end-to-end security solution for Software Defined Vehicular Networks (SDVN). Besides SDN's flexible network management, network performance, and energy-efficient end-toend security scheme plays a significant role in providing green IIoT services. Thus, the proposed SDVN provides lightweight end-to-end security. The end-to-end security objective is handled in two levels: i) In RSU-based Group Authentication (RGA) scheme, each vehicle in the RSU range receives a group id-key pair for secure communication and ii) In private-Collaborative Intrusion Detection System (p-CIDS), SDVN detects the potential intrusions inside the VANET architecture using collaborative learning that guarantees privacy through a fusion of differential privacy and homomorphic encryption schemes. The SDVN is simulated in NS2 & MATLAB, and results show increased energy efficiency with lower communication and storage overhead than existing frameworks. In addition, the p-CIDS detects the intruder with an accuracy of 96.81% in the SDVN.

show abstract

Section: Related Workmentioning

confidence: 99%

Energy-Efficient End-to-End Security for Software-Defined Vehicular Networks

Raja

Anbalagan

Vijayaraghavan

et al. 2021

IEEE Trans. Ind. Inf.

View full text Add to dashboard Cite

show abstract

“…To solve the optimization problem (20), f αk i (x) was set to f αk i (x) = 1 2x Tx + (b k i ) Tx for our approach in the simulations, where b k i was set to b k i = 1 k+1 c i + d i with c i ∈ R n and d i ∈ R n being constants private to agent i. Fig.…”

Section: A Evaluation Of Our Approachmentioning

confidence: 99%

“…One widely used approach to enabling privacypreservation in decentralized optimization is differential privacy [18]- [21] which protects sensitive information by adding carefully-designed noise to exchanged states or objective functions. However, adding noise also compromises the accuracy of optimization results and causes a fundamental trade-off between privacy and accuracy [18]- [20]. In fact, approaches based on differential privacy may fail to converge to the accurate optimization result even without noise perturbation [20].…”

Section: Introductionmentioning

confidence: 99%

“…However, adding noise also compromises the accuracy of optimization results and causes a fundamental trade-off between privacy and accuracy [18]- [20]. In fact, approaches based on differential privacy may fail to converge to the accurate optimization result even without noise perturbation [20]. It is worth noting that there exists some differential-privacy based optimization approaches which are able to converge to the accurate optimization result in the mean-square sense, e.g.…”

Section: Introductionmentioning

confidence: 99%

See 1 more Smart Citation

ADMM Based Privacy-Preserving Decentralized Optimization

Zhang

Ahmad

Wang

2019

IEEE Trans.Inform.Forensic Secur.

157

104

View full text Add to dashboard Cite

This paper considers the problem of privacypreservation in decentralized optimization, in which N agents cooperatively minimize a global objective function that is the sum of N local objective functions. We assume that each local objective function is private and only known to an individual agent. To cooperatively solve the problem, most existing decentralized optimization approaches require participating agents to exchange and disclose estimates to neighboring agents. However, this results in leakage of private information about local objective functions, which is undesirable when adversaries exist and try to steal information from participating agents. To address this issue, we propose a privacy-preserving decentralized optimization approach based on proximal Jacobian ADMM via function decomposition. Numerical simulations confirm the effectiveness of the proposed approach.

show abstract

“…While secure multi-party computation also deals with scenarios where no trust exists among agents, the maximum number of agents that can collude (without the privacy of others being breached) is bounded, whereas using differential privacy provides immunity against arbitrary collusions [Kairouz et al, 2015, Pettai andLaud, 2015]. As a result, differential privacy has been adopted by recent works in a number of areas pertaining to networked systems, such as control [Huang et al, 2012, estimation [Ny and Pappas, 2014], and optimization [Han et al, 2014, Huang et al, 2015, Nozari et al, 2017. Of relevance to our present work, the paper [Huang et al, 2012] studies the average consensus problem with differentially privacy guarantees and proposes an adjacency-based distributed algorithm with decaying Laplace noise and mean-square convergence.…”

Section: Introductionmentioning

confidence: 99%

Differentially private average consensus: Obstructions, trade-offs, and optimal algorithm design

2017

Self Cite

View full text Add to dashboard Cite

This paper studies the multi-agent average consensus problem under the requirement of differential privacy of the agents' initial states against an adversary that has access to all the messages. We first establish that a differentially private consensus algorithm cannot guarantee convergence of the agents' states to the exact average in distribution, which in turn implies the same impossibility for other stronger notions of convergence. This result motivates our design of a novel differentially private Laplacian consensus algorithm in which agents linearly perturb their state-transition and message-generating functions with exponentially decaying Laplace noise. We prove that our algorithm converges almost surely to an unbiased estimate of the average of agents' initial states, compute the exponential mean-square rate of convergence, and formally characterize its differential privacy properties. We show that the optimal choice of our design parameters (with respect to the variance of the convergence point around the exact average) corresponds to a one-shot perturbation of initial states and compare our design with various counterparts from the literature. Simulations illustrate our results.

show abstract

Differentially Private Distributed Convex Optimization via Functional Perturbation

Cited by 138 publications

References 26 publications

Energy-Efficient End-to-End Security for Software-Defined Vehicular Networks

Energy-Efficient End-to-End Security for Software-Defined Vehicular Networks

ADMM Based Privacy-Preserving Decentralized Optimization

Differentially private average consensus: Obstructions, trade-offs, and optimal algorithm design

Contact Info

Product

Resources

About