Privacy Attacks Research Articles

Preserving Location Privacy in the IoT against Advanced Attacks using Deep Learning

Location-based services (LBSs) have received a significant amount of recent attention from the research community due to their valuable benefits in various aspects of society. In addition, the dependency on LBS in the performance of daily tasks has increased dramatically, especially after the spread of the COVID-19 pandemic. LBS users use their real location to build LBS queries to take benefits. This makes location privacy vulnerable to attacks. The privacy issue is accentuated if the attacker is an LBS provider since all information about users is accessible. Moreover, the attacker can apply advanced attacks, such as map matching and semantic location attacks. In response to these issues, this work employs artificial intelligence to build a robust defense against advanced location privacy attacks. The key idea behind protecting the location privacy of LBS users is to generate smart dummy locations. Smart dummy locations are false locations with the same query probability as the real location, but they are far from both the real location and each other. Relying on the previous two conditions, the deep-learning-based intelligent finder ensures a high level of location privacy protection against advanced attacks. The attacker cannot recognize the dummies from the real location and cannot isolate the real location by a filtering process. In terms of entropy (the privacy protection metric), accuracy (the deep learning metric), and total execution time (the performance metric) and compared to the well-known DDA and BDA systems, the proposed system shows better results, where entropy = 15.9, accuracy = 9.9, and total execution time = 17 sec.

Cancelable ECG biometric based on combination of deep transfer learning with DNA and amino acid approaches for human authentication

Recently, electrocardiogram (ECG) signals have received a high level of attention as a physiological signal in the field of biometrics. It has presented great possibilities for its strength against counterfeit. However, the ECG feature templates are irreplaceable, and a compromised template implies a permanent loss of identity. Therefore, several studies have been introduced biometric template protection techniques such as cancelable techniques to protect the original template in case it is stolen or lost. In this research, a cancelable ECG approach is proposed to protect the ECG feature template for human authentication. In our system, we first employed some image processing techniques for preprocessing the input ECG signals. Then, a deep transfer learning approach is employed to extract the deep ECG features. Later, the proposed cancelable approach based on DNA and amino acid is applied to protect the deep feature templates. Lastly, a Support Vector Machine (SVM) is employed for authentication. Extensive experiments on two commonly used datasets coupled with comprehensive theoretical analysis demonstrate the highest accuracy of the proposed system and the strong resilience of the system to various security and privacy attacks. Results show that the proposed cancelable method meets all requirements of cancelable biometrics such as irreversibility, revocability, and unlinkability.

Topology-theoretic approach to address attribute linkage attacks in differential privacy

Differential Privacy (DP) is well-known for its strong privacy guarantee. Briefly speaking, DP algorithms guarantee that the statistical information of the data is roughly preserved, and at the same time, individual privacy is protected with guarantees. However, when there are correlations among the attribute in the dataset, only relying on DP is not sufficient to defend against the attribute linkage attack, which is a well-known privacy attack aiming at deducing individuals’ private information. In the attribute linkage attack, the adversary can leverage prior knowledge about the victim, combined with accessing the published dataset, to infer sensitive information about a victim. In this paper, we study the attribute linkage attack in DP settings, and argue that enhancing DP can give users a higher level of privacy guarantees. Our contributions are ➀ we show that the attribute linkage attack can be initiated with high probability under the protection of DP, ➁ we propose a variant of DP called APL-Freeϵ-DP to provide a higher level of privacy guarantees, ➂ we design an algorithm APLKiller which satisfies the APL-Free ϵ-DP. Finally, experiments show that our algorithm not only eliminates the attribute linkage attack, and at the same time, it has a better ability to extract useful information from the data.

Spears and shields: attacking and defending deep model co-inference in vehicular crowdsensing networks

Vehicular CrowdSensing (VCS) network is one of the key scenarios for future 6G ubiquitous artificial intelligence. In a VCS network, vehicles are recruited for collecting urban data and performing deep model inference. Due to the limited computing power of vehicles, we deploy a device-edge co-inference paradigm to improve the inference efficiency in the VCS network. Specifically, the vehicular device and the edge server keep a part of the deep model separately, but work together to perform the inference through sharing intermediate results. Although vehicles keep the raw data locally, privacy issues still exist once attackers obtain the shared intermediate results and recover the raw data in some way. In this paper, we validate the possibility by conducting a systematic study on the privacy attack and defense in the co-inference of VCS network. The main contributions are threefold: (1) We take the road sign classification task as an example to demonstrate how an attacker reconstructs the raw data without any knowledge of deep models. (2) We propose a model-perturbation defense to defend against such attacks by injecting some random Laplace noise into the deep model. A theoretical analysis is given to show that the proposed defense mechanism achieves epsilon-differential privacy. (3) We further propose a Stackelberg game-based incentive mechanism to attract the vehicles to participate in the co-inference by compensating their privacy loss in a satisfactory way. The simulation results show that our proposed defense mechanism can significantly reduce the effects of the attacks and the proposed incentive mechanism is very effective.

A Privacy Attack on Coded Caching by Colluding Users

Coded caching is beneficial due to its ability to leverage the multicast property of a wireless channel. Under specific network settings, the coded caching scheme has been shown to reveal the user file requests. In prior work, a privacy attack by a single attacker on coded caching was presented. In this letter, we focus on a relatively secure setting whereby only the intended users access the files. Nevertheless, we show that, even in this setting, privacy is compromised if the users are colluding. The time complexity of the attack on a network with $N$ files, $K$ users, and $l \leq K$ colluding users is shown to be in the order of $\sqrt {\ln (N(K-l))}N^{(K-l)/2}$ steps.

#PrettyFlyForAWiFi

Drones are becoming increasingly popular for hobbyists and recreational use. But with this surge in popularity comes increased risk to privacy as the technology makes it easy to spy on people in otherwise-private environments, such as an individual’s home. An attacker can fly a drone over fences and walls to observe the inside of a house, without having physical access. Existing drone detection systems require specialist hardware and expensive deployment efforts, making them inaccessible to the general public. In this work, we present a drone detection system that requires minimal prior configuration and uses inexpensive commercial off-the-shelf hardware to detect drones that are carrying out privacy invasion attacks. We use a model of the attack structure to derive statistical metrics for movement and proximity that are then applied to received communications between a drone and its controller. We test our system in real-world experiments with two popular consumer drone models mounting privacy invasion attacks using a range of flight patterns. We are able both to detect the presence of a drone and to identify which phase of the privacy attack was in progress while being resistant to false positives from other mobile transmitters. For line-of-sight approaches using our kurtosis-based method, we are able to detect all drones at a distance of 6 m, with the majority of approaches detected at 25 m or farther from the target window without suffering false positives for stationary or mobile non-drone transmitters.

Achieving security and privacy in federated learning systems: Survey, research challenges and future directions

Federated learning (FL) allows a server to learn a machine learning (ML) model across multiple decentralized clients that privately store their own training data. In contrast with centralized ML approaches, FL saves computation to the server and does not require the clients to outsource their private data to the server. However, FL is not free of issues. On the one hand, the model updates sent by the clients at each training epoch might leak information on the clients’ private data. On the other hand, the model learnt by the server may be subjected to attacks by malicious clients; these security attacks might poison the model or prevent it from converging. In this paper, we first examine security and privacy attacks to FL and critically survey solutions proposed in the literature to mitigate each attack. Afterwards, we discuss the difficulty of simultaneously achieving security and privacy protection. Finally, we sketch ways to tackle this open problem and attain both security and privacy.

Scrutinizing the Vulnerability of Ephemeral Diffie–Hellman over COSE (EDHOC) for IoT Environment Using Formal Approaches

Most existing conventional security mechanisms are insufficient, mainly attributable to their requirements for heavy processing capacity, large protocol message size, and longer round trips, for resource-intensive devices operating in an Internet of Things (IoT) context. These devices necessitate efficient communication and security protocols that are cognizant of the severe resource restrictions regarding energy, computation, communication, and storage. To realize this, the IETF (Internet Engineering Task Force) is currently working towards standardizing an ephemeral key-based lightweight and authenticated key exchange protocol called EDHOC (Ephemeral Diffie–Hellman over COSE). The protocol’s primary purpose is to build an OSCORE (Object Security for Constrained RESTful Environments) security environment by supplying crucial security properties such as secure key exchange, mutual authentication, perfect forward secrecy, and identity protection. EDHOC will most likely dominate IoT security once it becomes a standard. It is, therefore, imperative to inspect the protocol for any security flaw. In this regard, two previous studies have shown different security vulnerabilities of the protocol using formal security verification methods. Yet, both missed the vital security flaws we found in this paper: resource exhaustion and privacy attacks. In finding these vulnerabilities, we leveraged BAN-Logic and AVISPA to formally verify both EDHOC protocol variants. Consequently, we described these security flaws together with the results of the related studies and put forward recommended solutions as part of our future work.

Preserving User Privacy in Personalized Networks

The advances in efficient computation and data processing technologies have ushered in an entirely new era in which network connectivity applications are being revolutionized by Artificial Intelligence (AI). One of the emerging AI applications in wireless networks is network personalization. Personalized wireless networks are designed to track user satisfaction levels using Machine Learning (ML) models that are trained and updated using context data collected from users. Nonetheless, user data collection and transmission make sensitive user information prone to various privacy attacks. Therefore, in this letter, we propose a privacy-preserving framework for personalized wireless networks. The proposed framework is enabled by Differential Privacy (DP), which is an influential privacy technology that offers a rigorous and provable privacy guarantee. Besides, the proposed framework is supported by a multi-stage prediction process, which adds a second layer of privacy protection to personal user data. Lastly, we investigate the impact of privacy budget and data size on the quality of user satisfaction predictors in personalized wireless networks.

A novel road side unit assisted hash chain based approach for authentication in vehicular Ad-hoc network

Growth of connected vehicles in vehicular social network (VSN) for the ease of driving, route recommendation and infotainment services involve huge exchange of messages between the vehicles. This exchange of messages is prone to several security attacks. Also, the messages must be from authentic source for the success of VSN. Thus, VSN attracts industries, researchers and academia for the development and deployment of a network for its secured communication. Vehicular Ad-hoc Network (VANET) can provide a promising solution for connected vehicles. VANET is a framework in the intelligent transportation systems (ITS) which provide communication between vehicles also known as vehicle to vehicle (V2V) and to infrastructure also known as vehicle to infrastructure (V2I) communications using dedicated short range communication (DSRC) and wireless Access for Vehicular Environment (WAVE). Several security and privacy majors have been taken up by the researchers in the literature. Also most of the scheme completely relies on the trusted authority (TA) for all decisions. However, because of the strict time constraints, the delay must be minimum and incurs less computational and communication overhead. In this research work, RSU assisted hash chain based approach has been proposed to mitigate security and privacy attacks. RSU based scheme thus reduces the complete dependency on the TA. Simulation results show the effectiveness of proposed work as compared to existing work mentioned in this paper.

Privacy preservation in federated learning: An insightful survey from the GDPR perspective

In recent years, along with the blooming of Machine Learning (ML)-based applications and services, ensuring data privacy and security have become a critical obligation. ML-based service providers not only confront with difficulties in collecting and managing data across heterogeneous sources but also challenges of complying with rigorous data protection regulations such as EU/UK General Data Protection Regulation (GDPR). Furthermore, conventional centralised ML approaches have always come with long-standing privacy risks to personal data leakage, misuse, and abuse. Federated learning (FL) has emerged as a prospective solution that facilitates distributed collaborative learning without disclosing original training data. Unfortunately, retaining data and computation on-device as in FL are not sufficient for privacy-guarantee because model parameters exchanged among participants conceal sensitive information that can be exploited in privacy attacks. Consequently, FL-based systems are not naturally compliant with the GDPR. This article is dedicated to surveying of state-of-the-art privacy-preservation techniques in FL in relations with GDPR requirements. Furthermore, insights into the existing challenges are examined along with the prospective approaches following the GDPR regulatory guidelines that FL-based systems shall implement to fully comply with the GDPR.

Privacy Preserving Face Recognition in Cloud Robotics: A Comparative Study

Real-time robotic applications encounter the robot on board resources’ limitations. The speed of robot face recognition can be improved by incorporating cloud technology. However, the transmission of data to the cloud servers exposes the data to security and privacy attacks. Therefore, encryption algorithms need to be set up. This paper aims to study the security and performance of potential encryption algorithms and their impact on the deep-learning-based face recognition task’s accuracy. To this end, experiments are conducted for robot face recognition through various deep learning algorithms after encrypting the images of the ORL database using cryptography and image-processing based algorithms.

Anomaly detection in autonomous electric vehicles using AI techniques: A comprehensive survey

AbstractThe next wave in smart transportation is directed towards the design of renewable energy sources that can fuel automobile sector to shift towards the autonomous electric vehicles (AEVs). AEVs are sensor‐driven and driverless that uses artificial intelligence (AI)‐based interactions in Internet‐of‐vehicles (IoV) ecosystems. AEVs can reduce carbon footprints and trade energy with peer AEVs, smart grids (SG), and roadside units (RSUs). It supports green transportation vision. However, the sensor information, energy units, and user data are exchanged through open channels, and thus, are susceptible to various security and privacy attacks. Thus, AEVs can be remotely operated and directed by malicious entities that can propagate false updates to the peer nodes in IoV environment. This can cause the failure of components, congestion, as well as the entire disruption of IoV network. Globally researchers and security analysts have addressed solutions that pertain to specific security requirements, but still, the detection and classification of malicious AEVs is a widely studied topic. Malicious AEVs exhibit an anomaly behavior that differentiates them from normal AEVs, and thereby, the detection of anomalous AEVs and classification of anomaly type is required. Motivated from the aforementioned facts, the survey presents a systematic outlook of AI techniques in anomaly detection of AEVs. A solution taxonomy is proposed based on research gaps in the existing surveys, and the evaluation metrics for AI‐based anomaly detection are discussed. The open challenges and issues in AI deployments are discussed and a case study is presented on anomaly classification through a weighted ensemble technique. Thus, the proposed survey is designed to guide the manufacturing industry, AI practitioners, and researchers worldwide to formulate and design accurate and precise mechanisms to detect anomalies.

Methods of privacy-preserving genomic sequencing data alignments.

Genomic data alignment, a fundamental operation in sequencing, can be utilized to map reads into a reference sequence, query on a genomic database and perform genetic tests. However, with the reduction of sequencing cost and the accumulation of genome data, privacy-preserving genomic sequencing data alignment is becoming unprecedentedly important. In this paper, we present a comprehensive review of secure genomic data comparison schemes. We discuss the privacy threats, including adversaries and privacy attacks. The attacks can be categorized into inference, membership, identity tracing and completion attacks and have been applied to obtaining the genomic privacy information. We classify the state-of-the-art genomic privacy-preserving alignment methods into three different scenarios: large-scale reads mapping, encrypted genomic datasets querying and genetic testing to ease privacy threats. A comprehensive analysis of these approaches has been carried out to evaluate the computation and communication complexity as well as the privacy requirements. The survey provides the researchers with the current trends and the insights on the significance and challenges of privacy issues in genomic data alignment.

Modeling Deep Learning Based Privacy Attacks on Physical Mail

Mail privacy protection aims to prevent unauthorized access to hidden content within an envelope since normal paper envelopes are not as safe as we think. In this paper, for the first time, we show that with a well designed deep learning model, the hidden content may be largely recovered without opening the envelope. We start by modeling deep learning-based privacy attacks on physical mail content as learning the mapping from the camera-captured envelope front face image to the hidden content, then we explicitly model the mapping as a combination of perspective transformation, image dehazing and denoising using a deep convolutional neural network, named Neural-STE (See-Through-Envelope). We show experimentally that hidden content details, such as texture and image structure, can be clearly recovered. Finally, our formulation and model allow us to design envelopes that can counter deep learning-based privacy attacks on physical mail.

Improving Robustness to Model Inversion Attacks via Mutual Information Regularization

This paper studies defense mechanisms against model inversion (MI) attacks -- a type of privacy attacks aimed at inferring information about the training data distribution given the access to a target machine learning model. Existing defense mechanisms rely on model-specific heuristics or noise injection. While being able to mitigate attacks, existing methods significantly hinder model performance. There remains a question of how to design a defense mechanism that is applicable to a variety of models and achieves better utility-privacy tradeoff. In this paper, we propose the Mutual Information Regularization based Defense (MID) against MI attacks. The key idea is to limit the information about the model input contained in the prediction, thereby limiting the ability of an adversary to infer the private training attributes from the model prediction. Our defense principle is model-agnostic and we present tractable approximations to the regularizer for linear regression, decision trees, and neural networks, which have been successfully attacked by prior work if not attached with any defenses. We present a formal study of MI attacks by devising a rigorous game-based definition and quantifying the associated information leakage. Our theoretical analysis sheds light on the inefficacy of DP in defending against MI attacks, which has been empirically observed in several prior works. Our experiments demonstrate that MID leads to state-of-the-art performance for a variety of MI attacks, target models and datasets.

Efficient Certificate-Less Aggregate Signature Scheme with Conditional Privacy-Preservation for Vehicular Ad Hoc Networks Enhanced Smart Grid System.

Vehicular Ad hoc networks (VANETs) as spontaneous wireless communication technology of vehicles has a wide range of applications like road safety, navigation and other electric car technologies, however its practicability is greatly hampered by cyber-attacks. Due to message broadcasting in an open environment during communication, VANETs are inherently vulnerable to security and privacy attacks. However to address the cyber-security issues with optimal computation overhead is a matter of current security research challenge. So this paper designs a secure and efficient certificate-less aggregate scheme (ECLAS) for VANETs applicable in a smart grid scenario. The proposed scheme is based on elliptic curve cryptography to provide conditional privacy-preservation by incorporating usage of time validated pseudo-identification for communicating vehicles besides sorting out the KGC (Key Generation Center) escrow problem. The proposed scheme is comparatively more efficient to relevant related research work because it precludes expensive computation operations likes bilinear pairings as shown by the performance evaluation. Similarly, communication cost is within the ideal range to most related works while considering the security requirements of VANETs system applicable in a smart grid environment.

Challenges and New Directions in Securing Spectrum Access Systems

The spectrum access system (SAS) is being deployed as a key component of the emerging spectrum sharing paradigm to address the spectrum crunch facing the U.S. wireless industry. Ensuring security and privacy of this system against potential attacks is a task of paramount importance. In this article, we first introduce the SAS system, describing its three-tier access model, its functional architecture, and the spectrum management protocol. We then provide a comprehensive analysis of a variety of security and privacy attacks that an SAS is vulnerable to, and discuss their countermeasures. We identify key challenges, formalize threat models, and organize the discussion of SAS security into four categories: 1) SAS server security and privacy; 2) citizens broadband radio service device security; 3) security of environment sensing capability; and 4) communication protocol security. Finally, we suggest future research directions for spectrum management security.

Privacy protection of online social network users, against attribute inference attacks, through the use of a set of exhaustive rules

A malicious data miner can infer users’ private information in online social networks (OSNs) by data mining the users’ disclosed information. By exploring the public information about a target user (i.e. an individual or a group of OSN users whose privacy is under attack), attackers can prepare a training data set holding similar information about other users who openly disclosed their data. Using a machine learning classifier, the attacker can input released information about users under attack as non-class attributes and extract the private information as a class attribute. Some techniques offer some privacy protection against specific classifiers;, however, the provided privacy can be under threat if an attacker uses a different classifier (rather than the one used by the privacy protection techniques) to infer sensitive information. In reality, it is difficult to predict the classifiers involved in a privacy attack. In this study, we propose a privacy-preserving technique which first prepares a training data set in a similar way that an attacker can prepare and then takes an approach independent of the classifiers to extract patterns (or logic rules) from the training data set. Based on the extracted rule set, it then suggests the target users to hide some non-class attribute values and/or modify some friendship links for protecting their privacy. We apply our proposed technique on two OSN data sets containing users’ attribute values and their friendship links. For evaluating the performance of the proposed technique, we use conventional classifiers such as Na $$\ddot{\text {i}}$$ ve Bayes, Support Vector Machine and Random Forest on the privacy-protected data sets. The experimental results show that our proposed technique outperforms the existing privacy-preserving algorithms in terms of securing privacy while maintaining the data utility.

(p+, α, t)-Anonymity Technique Against Privacy Attacks

Privacy preservation is a major concern in current technology where enormous amounts of data are being collected and published for carrying out analysis. These data may contain sensitive information related to individual who owns them. If the data is published in their original form, they may lead to privacy disclosure which threats privacy requirements. Hence, the data should be anonymized before publishing so that it becomes challenging for intruders to obtain sensitive information by means of any privacy attack model. There are popular data anonymization techniques such as k-anonymity, l-diversity, p-sensitive k-anonymity, (l, m, d) anonymity, and t-closeness, which are vulnerable to different privacy attacks discussed in this paper. The proposed technique called (p+, α, t)-anonymity aims to anonymize the data in such a way that even though intruder has sufficient background knowledge on the target individual he will not be able to infer anything and breach private information. The anonymized data also provide sufficient data utility by allowing various data analytics to be performed.