The Death and Rebirth of Privacy-Preserving WiFi Fingerprint Localization with Paillier Encryption

Yang, Zheng; Järvinen, Kimmo

doi:10.1109/infocom.2018.8486221

Cited by 50 publications

(61 citation statements)

References 27 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Recently, several solutions, e.g. [13,12,24,22], have been proposed to protect the clients' location privacy in ILSs. However, only a few pieces of research (e.g.…”

Section: Introductionmentioning

confidence: 99%

Modeling Privacy in WiFi Fingerprinting Indoor Localization

Yang

Järvinen

2018

Provable Security

Self Cite

View full text Add to dashboard Cite

In this paper, we study privacy models for privacy-preserving Wifi fingerprint based indoor localization (PPIL) schemes. We show that many existing models are insufficient and make unrealistic assumptions regarding adversaries' power. To cover the state-of-the-art practical attacks, we propose the first formal security model which formulates the security goals of both client-side and server-side privacy beyond the curious-but-honest setting. In particular, our model considers various malicious behaviors such as exposing secrets of principles, choosing malicious Wifi fingerprints in location queries, and specifying the location area of a target client. Furthermore, we formulate the client-side privacy in an indistinguishability manner where an adversary is required to distinguish a client's real location from a random one. The server-side privacy requires that adversaries cannot generate a fabricate database which provides a similar function to the real database of the server. In particular, we formally define the similarity between databases with a ball approach that has not been formalized before. We show the validity and applicability of our model by applying it to analyze the security of an existing PPIL protocol.

show abstract

“…Recently, several solutions, e.g. [13,12,24,22], have been proposed to protect the clients' location privacy in ILSs. However, only a few pieces of research (e.g.…”

Section: Introductionmentioning

confidence: 99%

Modeling Privacy in WiFi Fingerprinting Indoor Localization

Yang

Järvinen

2018

Provable Security

Self Cite

View full text Add to dashboard Cite

show abstract

“…We provide detailed benchmarking results for our best-performing and most precise algorithms for ranges of Reference Points (RPs) and Wi-Fi Access Points (APs). In addition, we provide even more detailed benchmarks for the same practical setting as in [35] with M =505 RPs and N =241 APs. Our most efficient algorithm has an online run-time of 0.15 s on two commodity servers which is orders of magnitude faster than prior work.…”

Section: B Our Contributionsmentioning

confidence: 99%

“…Li et al [37] design algorithms for Privacy-Preserving Indoor Localization (PPIL) based on the k-Nearest Neighbors (k-NN) algorithm relying on Paillier encryption. Their work has several shortcomings: Yang and Järvinen [35] very recently presented an attack against this system that reveals the server's database under a realistic attack model. Our most efficient Euclidean distance-based PPIL protocol performs 121× faster with respect to the total run-time and 553× faster in the online phase than the protocol of [37] (see §VI-E for details).…”

Section: Related Workmentioning

confidence: 99%

“…Yang and Järvinen [35] present four high-level proposals for implementing secure PPIL: fully homomorphic encryption, Garbled Circuits (GCs), Paillier encryption only, and Paillier encryption combined with GCs. They conclude that while some of them are clearly unpractical (especially, fully homomorphic encryption), at least the one combining Paillier encryption with GCs should be feasible in practice.…”

Section: Related Workmentioning

confidence: 99%

“…V we show the benchmarking results of our algorithms for N =241 APs, M =505 RPs, and k=3 Nearest Neighbors. The setting is the same as in [35] and represents a typical size for a single building database. Examples on databases with similar sizes can be found, e.g., in [103,104,105].…”

Section: E Performance Of Privacy-preserving Indoor Localizationmentioning

confidence: 99%

See 2 more Smart Citations

PILOT: Practical Privacy-Preserving Indoor Localization Using OuTsourcing

Järvinen

Leppäkoski

Lohan

et al. 2019

2019 IEEE European Symposium on Security and Privacy (EuroS&P)

Self Cite

View full text Add to dashboard Cite

In the last decade, we observed a constantly growing number of Location-Based Services (LBSs) used in indoor environments, such as for targeted advertising in shopping malls or finding nearby friends. Although privacy-preserving LBSs were addressed in the literature, there was a lack of attention to the problem of enhancing privacy of indoor localization, i.e., the process of obtaining the users' locations indoors and, thus, a prerequisite for any indoor LBS.In this work we present PILOT, the first practically efficient solution for Privacy-Preserving Indoor Localization (PPIL) that was obtained by a synergy of the research areas indoor localization and applied cryptography. We design, implement, and evaluate protocols for Wi-Fi fingerprint-based PPIL that rely on 4 different distance metrics. To save energy and network bandwidth for the mobile end devices in PPIL, we securely outsource the computations to two non-colluding semi-honest parties. Our solution mixes different secure two-party computation protocols and we design size-and depth-optimized circuits for PPIL. We construct efficient circuit building blocks that are of independent interest: Single Instruction Multiple Data (SIMD) capable oblivious access to an array with low circuit depth and selection of the k-Nearest Neighbors with small circuit size. Additionally, we reduce Received Signal Strength (RSS) values from 8 bits to 4 bits without any significant accuracy reduction. Our most efficient PPIL protocol is 553x faster than that of Li et al. (INFOCOM'14) and 500x faster than that of Ziegeldorf et al. (WiSec'14). Our implementation on commodity hardware has practical run-times of less than 1 second even for the most accurate distance metrics that we consider, and it can process more than half a million PPIL queries per day. 448 2019 IEEE European Symposium on Security and Privacy (EuroS&P)

show abstract