Junwei Zhang

Graph neural networks (GNNs) are extensively employed in location-related scenarios, relying on aggregation to gather features from neighboring nodes based on edge weights. Features are closely bound to nodes’ locations and edge weights mirror distance correlations. In this sense, certain privacy concerns exist while providing location-based services if there is insufficient privacy protection. To this end, we propose a privacy-preserving and distance-aware data aggregation framework (PEDA) for GNNs. Specifically, PEDA achieves location privacy by combining circular-based positional coding with inner product functional encryption. Because of the masks in the codes, the decryption returns masked distances, preventing distance leakage. Following this, in order to protect feature privacy, we employ secret sharing. To preserve the collection strategy’s privacy, we implement an oblivious transfer for collecting the shared features. Additionally, we securely generate the adjacency matrix and aggregate features based on multi-party computation. Thorough security analysis and comprehensive evaluation demonstrate the privacy, feasibility and practicality of our approach. When compared to related works, PEDA offers four types of privacy, maintains distance awareness and feature utility, and allows for oblivious data collecting with little computational cost sacrifice.

Privacy-preserving federated learning framework has been widely applied in the Internet of Vehicles (IoV) scenario, to enhance the privacy of local models and sensitive vehicle data. However, as increasing cyberattacks are launched against both the RSU and server side, such as node compromising and data inference attacks, it is necessary to improve the robustness of uploading local updates and prevent servers from obtaining additional private information of vehicles in the two-layered IoV model, while existing schemes hardly meet all the above-mentioned security requirements. Therefore, we propose a robust and unlinkable privacy-preserving federated learning framework RUPT-FL, that tolerates dropout RSUs while disassociating local updates from the RSUs who gathered them. Based on packed secret sharing, secure multi-party computation, and homomorphic encryption techniques, we design a robust and lightweight multi-RSUs-aided local update upload protocol for vehicles, then construct an identity-oblivious update reconstruction and aggregation protocol under the two-server model. We conduct a series of experiments to demonstrate that the proposed framework not only improves the robustness and accuracy of federated learning against a semi-honest adversary, but also avoids expensive computational overhead for vehicles.

The widespread use of machine learning in location-related scenarios is propelling the rapid development of intelligent transportation. To assist users in making more informed travel plans, the demand for improving prediction accuracy is growing. Prior to model training, data cleaning is a common method used to eliminate redundant, erroneous and outlier samples. However, in intelligent transportation, there are serious issues with location awareness and privacy protection of existing data cleaning schemes. Therefore, we propose a location-aware and privacy-preserving data cleaning framework (PriSPA) which provides a cleaned dataset consisting of the samples from adopted data suppliers at qualified locations while ensuring the privacy of locations, spatial constraints and sensitive samples. We combine boolean secret sharing with XOR operations to make sure that it is possible to figure out whether a location complies with spatial constraints without leakage. More specifically, we ensure privacy using key agreement, secret sharing, authenticated encryption and random permutation. We seriously analyze the security of PriSPA and conduct comprehensive experiments to prove its security, effectiveness and efficiency. Based on the comparisons with the raw traffic forecasting framework, we observe that PriSPA improves the precision of the model with 17.6% -32.7% error reduction.

As the core subject of IoT applications, IoT devices have faced numerous security challenges. Especially for IoT devices deployed in remote or harsh environments, they are often unattended for long periods, making it difficult to share the sensing data and susceptible to potential physical attacks. While aerial assistance methods represented by unmanned aerial vehicles (UAVs) can solve the problem of data sharing at a low cost, it is necessary to establish a secure channel between ground control stations, UAVs, and IoT devices due to the sensitivity of the sensing data. Recently, Physical Unclonable Function (PUF) has been proven to provide unique identity identification for devices using its tamper-proof feature. In this paper, we propose a lightweight UAV-assisted authentication and key agreement protocol for unattended IoT devices, ensuring secure communication and physical tamper-proof requirements. However, our work does not stop there. We noticed that some existing PUF-based authentication schemes misunderstand the ability of PUF, which leads to these schemes cannot actually provide physical protection. We analyzed the security vulnerabilities of these schemes and proposed rules that should be followed when designing authentication protocols using PUF. In addition, for the first time, we put forward the formal definitions and proof methods for PUF in the formal proof of the security protocol, which avoided the unreasonable initial assumptions adopted in the proof of the existing schemes. We extended Mao-Boyd (MB) logic and comprehensively analyzed the proposed protocol. We also evaluate the performance of the proposed scheme, and the results show that the proposed scheme has certain advantages in communication and computation overhead compared with existing schemes.

Recently, some Unbiased Gradient Sampling-based (UGS) methods have been proposed to enhance the security and efficiency of federated learning through crafted unbiased random transformation and sampling, such as MinMax Sampling in SIGMOD ’22. In this paper, we propose a novel attack, GLAUS, to show that UGS is not as secure as claimed in these works and is still vulnerable to the gradient leakage attack (GLA). Specifically, we demonstrate an idea to approximately infer the gradient for GLA in the context of the UGS scenario where the real gradient is not available. Once the gradient is approximately obtained, the security of the UGS frameworks is downgraded to that of the original federated learning. The approximate gradient is refined by the following steps: 1) narrow the gradient searching range to the finite set; 2) obtain the magnitude of each gradient value approximately; 3) revise the gradient signs . Versus the failure of existing attacks, extensive experiments on six datasets show that our attack is effective in reconstructing private datapoints with pixel-wise accuracy on four network sizes and three image resolutions. Finally, we show how to defend against GLAUS while maintaining the high efficiency of UGS and only introducing an additional step to hide the sampled gradient indices.

A private intersection-sum (PIS) scheme considers the private computing problem of how parties jointly compute the sum of associated values in the set intersection. In scenarios such as electronic voting, corporate credit investigation, and ad conversions, private data are held by different parties. However, despite two-party PIS being well-developed in many previous works, its extended version, multi-party PIS, has rarely been discussed thus far. This is because, depending on the existing works, directly initiating multiple two-party PIS instances is considered to be a straightforward way to achieve multi-party PIS; however, by using this approach, the intersection-sum results of the two parties and the data only belonging to the two-party intersection will be leaked. Therefore, achieving secure multi-party PIS is still a challenge. In this paper, we propose a secure and lightweight multi-party private intersection-sum scheme called SLMP-PIS. We maintain data privacy based on zero sharing and oblivious pseudorandom functions to compute the multi-party intersection and consider the privacy of associated values using arithmetic sharing and symmetric encryption. The security analysis results show that our protocol is proven to be secure in the standard semi-honest security model. In addition, the experiment results demonstrate that our scheme is efficient and feasible in practice. Specifically, when the number of participants is five, the efficiency can be increased by 22.98%.