Takao Murakami

Ting-Wei Liao, Chih-Hsun Lin, Yu-Lin Tsai et al.

Local Differential Privacy (LDP) has been widely adopted to protect user privacy in decentralized data collection. However, recent studies have revealed that LDP protocols are vulnerable to data poisoning attacks, where malicious users manipulate their reported data to distort aggregated results. In this work, we present the first study on data poisoning attacks targeting LDP range query protocols, focusing on both tree-based and grid-based approaches. We identify three key challenges in executing such attacks, including crafting consistent and effective fake data, maintaining data consistency across levels or grids, and preventing server detection. To address the first two challenges, we propose novel attack methods that are provably optimal, including a tree-based attack and a grid-based attack, designed to manipulate range query results with high effectiveness. \textbf{Our key finding is that the common post-processing procedure, Norm-Sub, in LDP range query protocols can help the attacker massively amplify their attack effectiveness.} In addition, we study a potential countermeasure, but also propose an adaptive attack capable of evading this defense to address the third challenge. We evaluate our methods through theoretical analysis and extensive experiments on synthetic and real-world datasets. Our results show that the proposed attacks can significantly amplify estimations for arbitrary range queries by manipulating a small fraction of users, providing 5-10x more influence than a normal user to the estimation.

Seira Hidano, Takao Murakami

Differentially private GNNs (Graph Neural Networks) have been recently studied to provide high accuracy in various tasks on graph data while strongly protecting user privacy. In particular, a recent study proposes an algorithm to protect each user's feature vector in an attributed graph, which includes feature vectors along with node IDs and edges, with LDP (Local Differential Privacy), a strong privacy notion without a trusted third party. However, this algorithm does not protect edges (friendships) in a social graph, hence cannot protect user privacy in unattributed graphs, which include only node IDs and edges. How to provide strong privacy with high accuracy in unattributed graphs remains open. In this paper, we propose a novel LDP algorithm called the DPRR (Degree-Preserving Randomized Response) to provide LDP for edges in GNNs. Our DPRR preserves each user's degree hence a graph structure while providing edge LDP. Technically, our DPRR uses Warner's RR (Randomized Response) and strategic edge sampling, where each user's sampling probability is automatically tuned using the Laplacian mechanism to preserve the degree information under edge LDP. We also propose a privacy budget allocation method to make the noise in both Warner's RR and the Laplacian mechanism small. We focus on graph classification as a task of GNNs and evaluate the DPRR using three social graph datasets. Our experimental results show that the DPRR significantly outperforms three baselines and provides accuracy close to a non-private algorithm in all datasets with a reasonable privacy budget, e.g., epsilon=1. Finally, we introduce data poisoning attacks to our DPRR and a defense against the attacks. We evaluate them using the three social graph datasets and discuss the experimental results.

Jacob Imola, Takao Murakami, Kamalika Chaudhuri

Triangle counting in networks under LDP (Local Differential Privacy) is a fundamental task for analyzing connection patterns or calculating a clustering coefficient while strongly protecting sensitive friendships from a central server. In particular, a recent study proposes an algorithm for this task that uses two rounds of interaction between users and the server to significantly reduce estimation error. However, this algorithm suffers from a prohibitively high communication cost due to a large noisy graph each user needs to download. In this work, we propose triangle counting algorithms under LDP with a small estimation error and communication cost. We first propose two-rounds algorithms consisting of edge sampling and carefully selecting edges each user downloads so that the estimation error is small. Then we propose a double clipping technique, which clips the number of edges and then the number of noisy triangles, to significantly reduce the sensitivity of each user's query. Through comprehensive evaluation, we show that our algorithms dramatically reduce the communication cost of the existing algorithm, e.g., from 6 hours to 8 seconds or less at a 20 Mbps download rate, while keeping a small estimation error.

Takao Murakami, Hiromi Arai, Koki Hamada et al.

For a better understanding of anonymization methods for location traces, we have designed and held a location trace anonymization contest that deals with a long trace (400 events per user) and fine-grained locations (1024 regions). In our contest, each team anonymizes her original traces, and then the other teams perform privacy attacks against the anonymized traces. In other words, both defense and attack compete together, which is close to what happens in real life. Prior to our contest, we show that re-identification alone is insufficient as a privacy risk and that trace inference should be added as an additional risk. Specifically, we show an example of anonymization that is perfectly secure against re-identification and is not secure against trace inference. Based on this, our contest evaluates both the re-identification risk and trace inference risk and analyzes their relationship. Through our contest, we show several findings in a situation where both defense and attack compete together. In particular, we show that an anonymization method secure against trace inference is also secure against re-identification under the presence of appropriate pseudonymization. We also report defense and attack algorithms that won first place, and analyze the utility of anonymized traces submitted by teams in various applications such as POI recommendation and geo-data analysis.

Seira Hidano, Takao Murakami, Yusuke Kawamoto

Transfer learning has been widely studied and gained increasing popularity to improve the accuracy of machine learning models by transferring some knowledge acquired in different training. However, no prior work has pointed out that transfer learning can strengthen privacy attacks on machine learning models. In this paper, we propose TransMIA (Transfer learning-based Membership Inference Attacks), which use transfer learning to perform membership inference attacks on the source model when the adversary is able to access the parameters of the transferred model. In particular, we propose a transfer shadow training technique, where an adversary employs the parameters of the transferred model to construct shadow models, to significantly improve the performance of membership inference when a limited amount of shadow training data is available to the adversary. We evaluate our attacks using two real datasets, and show that our attacks outperform the state-of-the-art that does not use our transfer shadow training technique. We also compare four combinations of the learning-based/entropy-based approach and the fine-tuning/freezing approach, all of which employ our transfer shadow training technique. Then we examine the performance of these four approaches based on the distributions of confidence values, and discuss possible countermeasures against our attacks.

Natasha Fernandes, Yusuke Kawamoto, Takao Murakami

Extended differential privacy, a generalization of standard differential privacy (DP) using a general metric, has been widely studied to provide rigorous privacy guarantees while keeping high utility. However, existing works on extended DP are limited to few metrics, such as the Euclidean metric. Consequently, they have only a small number of applications, such as location-based services and document processing. In this paper, we propose a couple of mechanisms providing extended DP with a different metric: angular distance (or cosine distance). Our mechanisms are based on locality sensitive hashing (LSH), which can be applied to the angular distance and work well for personal data in a high-dimensional space. We theoretically analyze the privacy properties of our mechanisms, and prove extended DP for input data by taking into account that LSH preserves the original metric only approximately. We apply our mechanisms to friend matching based on high-dimensional personal data with angular distance in the local model, and evaluate our mechanisms using two real datasets. We show that LDP requires a very large privacy budget and that RAPPOR does not work in this application. Then we show that our mechanisms enable friend matching with high utility and rigorous privacy guarantees based on extended DP.

Jacob Imola, Takao Murakami, Kamalika Chaudhuri

Differentially private analysis of graphs is widely used for releasing statistics from sensitive graphs while still preserving user privacy. Most existing algorithms however are in a centralized privacy model, where a trusted data curator holds the entire graph. As this model raises a number of privacy and security issues -- such as, the trustworthiness of the curator and the possibility of data breaches, it is desirable to consider algorithms in a more decentralized local model where no server holds the entire graph. In this work, we consider a local model, and present algorithms for counting subgraphs -- a fundamental task for analyzing the connection patterns in a graph -- with LDP (Local Differential Privacy). For triangle counts, we present algorithms that use one and two rounds of interaction, and show that an additional round can significantly improve the utility. For $k$-star counts, we present an algorithm that achieves an order optimal estimation error in the non-interactive local model. We provide new lower-bounds on the estimation error for general graph statistics including triangle counts and $k$-star counts. Finally, we perform extensive experiments on two real datasets, and show that it is indeed possible to accurately estimate subgraph counts in the local differential privacy model.

Takao Murakami, Kenta Takahashi

LDP (Local Differential Privacy) has recently attracted much attention as a metric of data privacy that prevents the inference of personal data from obfuscated data in the local model. However, there are scenarios in which the adversary wants to perform re-identification attacks to link the obfuscated data to users in this model. LDP can cause excessive obfuscation and destroy the utility in these scenarios because it is not designed to directly prevent re-identification. In this paper, we propose a measure of re-identification risks, which we call PIE (Personal Information Entropy). The PIE is designed so that it directly prevents re-identification attacks in the local model. It lower-bounds the lowest possible re-identification error probability (i.e., Bayes error probability) of the adversary. We analyze the relation between LDP and the PIE, and analyze the PIE and utility in distribution estimation for two obfuscation mechanisms providing LDP. Through experiments, we show that when we consider re-identification as a privacy risk, LDP can cause excessive obfuscation and destroy the utility. Then we show that the PIE can be used to guarantee low re-identification risks for the local obfuscation mechanisms while keeping high utility.

Takao Murakami, Koki Hamada, Yusuke Kawamoto et al.

With the widespread use of LBSs (Location-based Services), synthesizing location traces plays an increasingly important role in analyzing spatial big data while protecting user privacy. In particular, a synthetic trace that preserves a feature specific to a cluster of users (e.g., those who commute by train, those who go shopping) is important for various geo-data analysis tasks and for providing a synthetic location dataset. Although location synthesizers have been widely studied, existing synthesizers do not provide sufficient utility, privacy, or scalability, hence are not practical for large-scale location traces. To overcome this issue, we propose a novel location synthesizer called PPMTF (Privacy-Preserving Multiple Tensor Factorization). We model various statistical features of the original traces by a transition-count tensor and a visit-count tensor. We factorize these two tensors simultaneously via multiple tensor factorization, and train factor matrices via posterior sampling. Then we synthesize traces from reconstructed tensors, and perform a plausible deniability test for a synthetic trace. We comprehensively evaluate PPMTF using two datasets. Our experimental results show that PPMTF preserves various statistical features including cluster-specific features, protects user privacy, and synthesizes large-scale location traces in practical time. PPMTF also significantly outperforms the state-of-the-art methods in terms of utility and scalability at the same level of privacy.

Yusuke Kawamoto, Takao Murakami

We introduce a general model for the local obfuscation of probability distributions by probabilistic perturbation, e.g., by adding differentially private noise, and investigate its theoretical properties. Specifically, we relax a notion of distribution privacy (DistP) by generalizing it to divergence, and propose local obfuscation mechanisms that provide divergence distribution privacy. To provide f-divergence distribution privacy, we prove that probabilistic perturbation noise should be added proportionally to the Earth mover's distance between the probability distributions that we want to make indistinguishable. Furthermore, we introduce a local obfuscation mechanism, which we call a coupling mechanism, that provides divergence distribution privacy while optimizing the utility of obfuscated data by using exact/approximate auxiliary information on the input distributions we want to protect.

Yusuke Kawamoto, Takao Murakami

We introduce a formal model for the information leakage of probability distributions and define a notion called distribution privacy as the local differential privacy for probability distributions. Roughly, the distribution privacy of a local obfuscation mechanism means that the attacker cannot significantly gain any information on the distribution of the mechanism's input by observing its output. Then we show that existing local mechanisms can hide input distributions in terms of distribution privacy, while deteriorating the utility by adding too much noise. For example, we prove that the Laplace mechanism needs to add a large amount of noise proportionally to the infinite Wasserstein distance between the two distributions we want to make indistinguishable. To improve the tradeoff between distribution privacy and utility, we introduce a local obfuscation mechanism, called a tupling mechanism, that adds random dummy data to the output. Then we apply this mechanism to the protection of user attributes in location based services. By experiments, we demonstrate that the tupling mechanism outperforms popular local mechanisms in terms of attribute obfuscation and service quality.

Takao Murakami, Yusuke Kawamoto

LDP (Local Differential Privacy) has been widely studied to estimate statistics of personal data (e.g., distribution underlying the data) while protecting users' privacy. Although LDP does not require a trusted third party, it regards all personal data equally sensitive, which causes excessive obfuscation hence the loss of utility. In this paper, we introduce the notion of ULDP (Utility-optimized LDP), which provides a privacy guarantee equivalent to LDP only for sensitive data. We first consider the setting where all users use the same obfuscation mechanism, and propose two mechanisms providing ULDP: utility-optimized randomized response and utility-optimized RAPPOR. We then consider the setting where the distinction between sensitive and non-sensitive data can be different from user to user. For this setting, we propose a personalized ULDP mechanism with semantic tags to estimate the distribution of personal data with high utility while keeping secret what is sensitive for each user. We show theoretically and experimentally that our mechanisms provide much higher utility than the existing LDP mechanisms when there are a lot of non-sensitive data. We also show that when most of the data are non-sensitive, our mechanisms even provide almost the same utility as non-private mechanisms in the low privacy regime.

Yusuke Kawamoto, Takao Murakami

Obfuscation techniques in location-based services (LBSs) have been shown useful to hide the concrete locations of service users, whereas they do not necessarily provide the anonymity. We quantify the anonymity of the location data obfuscated by the planar Laplacian mechanism and that by the optimal geo-indistinguishable mechanism of Bordenabe et al. We empirically show that the latter provides stronger anonymity than the former in the sense that more users in the database satisfy k-anonymity. To formalize and analyze such approximate anonymity we introduce the notion of asymptotic anonymity. Then we show that the location data obfuscated by the optimal geo-indistinguishable mechanism can be anonymized by removing a smaller number of users from the database. Furthermore, we demonstrate that the optimal geo-indistinguishable mechanism has better utility both for users and for data analysts.

Takao Murakami, Tetsushi Ohki, Yosuke Kaga et al.

A cancelable biometric scheme called correlation-invariant random filtering (CIRF) is known as a promising template protection scheme. This scheme transforms a biometric feature represented as an image via the 2D number theoretic transform (NTT) and random filtering. CIRF has perfect secrecy in that the transformed feature leaks no information about the original feature. However, CIRF cannot be applied to large-scale biometric identification, since the 2D inverse NTT in the matching phase requires high computational time. Furthermore, existing biometric indexing schemes cannot be used in conjunction with template protection schemes to speed up biometric identification, since a biometric index leaks some information about the original feature. In this paper, we propose a novel indexing scheme called "cancelable indexing" to speed up CIRF without losing its security properties. The proposed scheme is based on fast computation of CIRF via low-rank approximation of biometric images and via a minimum spanning tree representation of low-rank matrices in the Fourier domain. We prove that the transformed index leaks no information about the original index and the original biometric feature (i.e., perfect secrecy), and thoroughly discuss the security of the proposed scheme. We also demonstrate that it significantly reduces the one-to-many matching time using a finger-vein dataset that includes six fingers from 505 subjects.