Search

Summary: Knowledge graphs contain a large amount of entity and relational data, and graph neural networks, as a class of efficient graph representation techniques based on deep learning, excel in knowledge graph modeling. However, previous neural network architectures for the most part only learn node representations and do not fully consider the heterogeneity of data. In this article, we innovatively propose a privacy attack framework based on IoT, PAFI, which is able to classify entities and relations, learn embedding representations in multi‐relational graphs, and can be applied to some existing neural network algorithms. Based on this, a fine‐grained privacy attack model, FPM, is proposed, which can perform attack operations on multiple targets, achieve selectivity of target tasks, and greatly improve the generalization ability of the attack model. In this article, the effectiveness of PAFI and FPM is demonstrated by real network datasets, and compared with previous attack methods, both of which achieve good results. [ABSTRACT FROM AUTHOR]

Most studies on privacy in machine learning have primarily focused on supervised learning, with little research on privacy concerns in reinforcement learning. However, our study has demonstrated that observation information can be extracted through trajectory analysis. In this paper, we propose a variable information inference attack targeting the observation space of policy models, which is categorised into two types: observed value inference attack and observed variable inference. Our algorithm has demonstrated a high success rate in privacy inference attacks for both types of observation information. [ABSTRACT FROM AUTHOR]

Virtual reality is immersive, personalized, and highly interactive. The data is rich and diverse, and privacy involves a wide range. It is crucial to study the privacy protection of virtual reality. This paper introduces the development process of virtual reality privacy protection in detail. From the form of data generation, the virtual reality privacy protection system is proposed, divided into perception, transmission, application, and interaction. According to the shortcomings of current research, the pending problems and possible solutions faced in virtual reality privacy protection are summarized. [ABSTRACT FROM AUTHOR]

Abstract Modern automobiles are equipped with connectivity features to enhance the user’s comfort. Bluetooth is one such communication technology that is used to pair a personal device with an automotive infotainment unit. Upon pairing, the user could access the personal information on the phone through the automotive head unit with minimum distraction while driving. However, such connectivity introduces a possibility for privacy attacks. Hence, performing an in-depth analysis of the system with privacy constraints is extremely important to prevent unauthorized access to personal information. In this work, we perform a systematic analysis of the Bluetooth network of an automotive infotainment unit to exploit security and privacy-related vulnerabilities. We model the identified threat with respect to privacy constraints of the system, emphasize the severity of attacks through a standardized rating metric and then provide potential countermeasures to prevent the attack. We perform System Theoretic Process Analysis for Privacy as a part of the systematic analysis and use the Common Vulnerability Scoring System to derive attack severity. The identified vulnerabilities are due to design flaws and assumptions on Bluetooth protocol implementation on automotive infotainment systems. We then elicit the vulnerability by performing a privacy attack on the Automotive system in an actual vehicle. We use Android Open-Source Project to report our findings and propose defense strategies.

Modern automobiles are equipped with connectivity features to enhance the user's comfort. Bluetooth is one such communication technology that is used to pair a personal device with an automotive infotainment unit. Upon pairing, the user could access the personal information on the phone through the automotive head unit with minimum distraction while driving. However, such connectivity introduces a possibility for privacy attacks. Hence, performing an in-depth analysis of the system with privacy constraints is extremely important to prevent unauthorized access to personal information. In this work, we perform a systematic analysis of the Bluetooth network of an automotive infotainment unit to exploit security and privacy-related vulnerabilities. We model the identified threat with respect to privacy constraints of the system, emphasize the severity of attacks through a standardized rating metric and then provide potential countermeasures to prevent the attack. We perform System Theoretic Process Analysis for Privacy as a part of the systematic analysis and use the Common Vulnerability Scoring System to derive attack severity. The identified vulnerabilities are due to design flaws and assumptions on Bluetooth protocol implementation on automotive infotainment systems. We then elicit the vulnerability by performing a privacy attack on the Automotive system in an actual vehicle. We use Android Open-Source Project to report our findings and propose defense strategies. [ABSTRACT FROM AUTHOR]

Bluetooth Low Energy (BLE) has become the de facto communication protocol for the Internet of Things (IoT) and smart wearable devices for its ultra-low energy consumption, ease of development, good enough network coverage, and data transfer speed. Due to the simplified design of this protocol, there have been lots of security and privacy vulnerabilities. As billions of health care, personal fitness wearable, smart lock, industrial automation devices adopt this technology for communication, its vulnerabilities should be dealt with high priority. Some segregated works on BLE were performed focusing on various vulnerabilities, such as the insecure implementation of encryption, device authentication, user privacy, etc. However, there has been no comprehensive survey on the security vulnerabilities of this protocol. In this survey paper, we present a comprehensive taxonomy for the security and privacy issues of BLE. We present possible attack scenarios for different types of vulnerabilities, classify them according to their severity, and list possible mitigation techniques. We also provide case studies regarding how different vulnerabilities can be exploited in real BLE devices.

Federated learning facilitates collaborative learning among multiple parties while ensuring client privacy. The vulnerability of federated learning to diverse Byzantine attacks stems from the opacity inherent in its local training processes. This issue becomes more pronounced when data among clients lacks independent and identical distribution. Concurrently, the advent of gradient inversion attacks introduces an escalating risk to the confidentiality and integrity of the federated learning system. This work introduces a framework to tackle the difficulties associated with Non-IID data within federated learning. Initially, it classifies diverse data using Euclidean distance to alleviate intra-cluster heterogeneity. A privacy-preserving Byzantine fault tolerance strategy utilizing cosine similarity is implemented within each cluster. We adopt normalizing and introduce historical momentum to the inter-cluster aggregation. The integration of homomorphic encryption ensures that clients' gradient information participates in training under ciphertext, safeguarding against gradient inversion attacks. Lastly, we conduct a comparative analysis against classical and latest algorithms under various conditions. These results validate the effectiveness of our design in enhancing robustness in Non-IID data. [ABSTRACT FROM AUTHOR]

The widespread adoption of deep learning is facilitated in part by the availability of large-scale data for training desirable models. However, these data may involve sensitive personal information, which raises privacy concerns for data providers. Differential privacy has been thought of as a key technique in the privacy preservation field, which has drawn much attention owing to its capability of providing rigorous and provable privacy guarantees for training data. Training deep learning models in a differentially private manner is a topic that is gaining traction as this alleviates the reconstruction and inference of sensitive information effectively. Taking this cue, in this paper, we present here a comprehensive and systematic study on differentially private deep learning from the facets of privacy attack and privacy preservation. We explore a new taxonomy to analyze the privacy attacks faced in deep learning and then survey the type of privacy preservation based on differential privacy to tackle such privacy attacks in deep learning. Finally, we propose the first probe into the real-world application of differentially private deep learning, and then conclude with several potential future research avenues. This survey provides promising directions for protecting sensitive information in training data via differential privacy during deep learning model training. [ABSTRACT FROM AUTHOR]

Privacy regulations and the physical distribution of heterogeneous data are often primary concerns for the development of deep learning models in a medical context. This paper evaluates the feasibility of differentially private federated learning for chest X-ray classification as a defense against data privacy attacks. To the best of our knowledge, we are the first to directly compare the impact of differentially private training on two different neural network architectures, DenseNet121 and ResNet50. Extending the federated learning environments previously analyzed in terms of privacy, we simulated a heterogeneous and imbalanced federated setting by distributing images from the public CheXpert and Mendeley chest X-ray datasets unevenly among 36 clients. Both non-private baseline models achieved an area under the receiver operating characteristic curve (AUC) of 0.94 on the binary classification task of detecting the presence of a medical finding. We demonstrate that both model architectures are vulnerable to privacy violation by applying image reconstruction attacks to local model updates from individual clients. The attack was particularly successful during later training stages. To mitigate the risk of a privacy breach, we integrated Rényi differential privacy with a Gaussian noise mechanism into local model training. We evaluate model performance and attack vulnerability for privacy budgets ε ∈ { 1 , 3 , 6 , 10 } . The DenseNet121 achieved the best utility-privacy trade-off with an AUC of 0.94 for ε = 6 . Model performance deteriorated slightly for individual clients compared to the non-private baseline. The ResNet50 only reached an AUC of 0.76 in the same privacy setting. Its performance was inferior to that of the DenseNet121 for all considered privacy constraints, suggesting that the DenseNet121 architecture is more robust to differentially private training. [ABSTRACT FROM AUTHOR]

Privacy regulations and the physical distribution of heterogeneous data are often primary concerns for the development of deep learning models in a medical context. This paper evaluates the feasibility of differentially private federated learning for chest X-ray classification as a defense against data privacy attacks. To the best of our knowledge, we are the first to directly compare the impact of differentially private training on two different neural network architectures, DenseNet121 and ResNet50. Extending the federated learning environments previously analyzed in terms of privacy, we simulated a heterogeneous and imbalanced federated setting by distributing images from the public CheXpert and Mendeley chest X-ray datasets unevenly among 36 clients. Both non-private baseline models achieved an area under the receiver operating characteristic curve (AUC) of 0.94 on the binary classification task of detecting the presence of a medical finding. We demonstrate that both model architectures are vulnerable to privacy violation by applying image reconstruction attacks to local model updates from individual clients. The attack was particularly successful during later training stages. To mitigate the risk of a privacy breach, we integrated Rényi differential privacy with a Gaussian noise mechanism into local model training. We evaluate model performance and attack vulnerability for privacy budgets ε∈{1,3,6,10}. The DenseNet121 achieved the best utility-privacy trade-off with an AUC of 0.94 for ε=6. Model performance deteriorated slightly for individual clients compared to the non-private baseline. The ResNet50 only reached an AUC of 0.76 in the same privacy setting. Its performance was inferior to that of the DenseNet121 for all considered privacy constraints, suggesting that the DenseNet121 architecture is more robust to differentially private training.

Being able to identify records that correspond to the same entity across diverse databases is an increasingly important step in many data analytics projects. Research into privacy-preserving record linkage (PPRL) aims to develop techniques that can link records across databases such that besides the record pairs classified as matches no sensitive information about the entities in these databases is revealed. A popular technique used in PPRL is to encode sensitive values into Bloom filters (bit vectors), which has the advantage of allowing approximate matching using character q-grams. PPRL based on Bloom filter encoding has been shown to be accurate and scalable to large databases, and is thus now being used in real-world PPRL systems in Australia, Canada, and the UK. However, recent studies have shown that Bloom filters used for PPRL are vulnerable to cryptanalysis attacks that can re-identify some of the sensitive values encoded in these Bloom filters. While previous such attack methods were slow and required knowledge of various encoding parameters, we present a novel efficient attack which exploits how attribute values are encoded into Bloom filters. Our attack method does not require knowledge of the encoding function or its parameter settings used. It is able to correctly re-identify with high precision q-grams that could not have been hashed to certain Bloom filter bit positions, and using these re-identified q-grams it can then re-identify attribute values with high precision. Our method is significantly faster than earlier PPRL cryptanalysis attacks, and in our experimental evaluation, it is able to successfully re-identify attribute values from large real-world databases in a few minutes. [ABSTRACT FROM AUTHOR]

We propose new privacy attacks to infer attributes (e.g., locations, occupations, and interests) of online social network users. Our attacks leverage seemingly innocent user information that is publicly available in online social networks to infer missing attributes of targeted users. Given the increasing availability of (seemingly innocent) user information online, our results have serious implications for Internet privacy--private attributes can be inferred from users' publicly available data unless we take steps to protect users from such inference attacks. To infer attributes of a targeted user, existing inference attacks leverage either the user's publicly available social friends or the user's behavioral records (e.g., the web pages that the user has liked on Facebook, the apps that the user has reviewed on Google Play), but not both. As we will show, such inference attacks achieve limited success rates. However, the problem becomes qualitatively different if we consider both social friends and behavioral records. To address this challenge, we develop a novel model to integrate social friends and behavioral records, and design new attacks based on our model. We theoretically and experimentally demonstrate the effectiveness of our attacks. For instance, we observe that, in a real-world large-scale dataset with 1.1 million users, our attack can correctly infer the cities a user lived in for 57% of the users; via confidence estimation, we are able to increase the attack success rate to over 90% if the attacker selectively attacks half of the users. Moreover, we show that our attack can correctly infer attributes for significantly more users than previous attacks. [ABSTRACT FROM AUTHOR]

Trajectory data often provide useful information that can be used in real-life applications, such as traffic management, Geo-marketing, and location-based advertising. However, a trajectory database may contain detailed information about moving objects and associate them with sensitive attributes, such as disease, job, and income. Therefore, improper publishing of the trajectory database can put the privacy of moving objects at risk, especially when an adversary uses partial trajectory information as its background knowledge. The existing approaches for privacy preservation in trajectory data publishing provide the same privacy protection for all moving objects. The consequence is that some moving objects may be offered insufficient privacy protection, while some others may not require high privacy protection. In this paper, we address this problem and present PPTD, a novel approach for preserving privacy in trajectory data publishing based on the concept of personalized privacy. It aims to strike a balance between the conflicting goals of data utility and data privacy in accordance with the privacy requirements of moving objects. To the best of our knowledge, this is the first paper that combines sensitive attribute generalization and trajectory local suppression to achieve a tailored personalized privacy model for trajectory data publishing. Our experiments on two synthetic trajectory datasets suggest that PPTD is effective for preserving personalized privacy in trajectory data publishing. In particular, PPTD can significantly improve the data utility of anonymized trajectory databases when compared with previous work in the literature. [ABSTRACT FROM AUTHOR]

Privacy-preserving record linkage (PPRL) is the process of identifying records in sensitive databases that refer to the same entities in applications where no private or confidential data can be shared by the owners of the databases being linked. In their paper "Blockchain-based Privacy-Preserving Record Linkage — Enhancing Data Privacy in an Untrusted Environment" (Nóbrega et al., 2021) (named BC-PPRL in the following), Nóbrega et al. (2021) proposed the use of blockchain technologies to provide accountability of the parties involved in a PPRL protocol and thereby allow the detection of misbehaving parties. While the use of blockchain techniques is an interesting and novel contribution to the research area of PPRL, as we show in this paper both theoretically and practically using a simple attack method, the BC-PPRL approach has some serious privacy weaknesses. We specifically highlight that one key aspect of the proposed approach, the exchange of Bloom filter segments between the database owners, can reveal substantially more sensitive information compared to what is stated in the paper by Nóbrega et al. (2021). Using a real-world data set we show how our attack can allow a database owner to reidentify with high accuracy a large number of the sensitive values that were encoded in the Bloom filter segments they receive from another database owner. We make the code and data sets of our attack available at: https://github.com/anushkavidanage/bc-pprlSegmentAtomAttack/. [ABSTRACT FROM AUTHOR]