Voiceprint verification has been widely adopted in daily applications due to its convenience and efficiency. However, the latent representations extracted during speech verification often encode sensitive personal information, such as emotion and gender, which raises significant privacy concerns. Existing methods mainly focus on protecting predefined sensitive attributes, but in many cases, the attributes that attackers are interested in are unknown to defenders. To address this limitation, we propose the Speech Information Bottleneck Network (SIBNet), a framework for representation level privacy enhancement based on information theoretic approaches. SIBNet differs from existing speech attribute privacy mechanisms in three key respects: (1) We formulate biometric privacy protection as an Information Bottleneck (IB) optimization problem that balances identity utility and sensitive attribute suppression. (2) SIBNet leverages a variational distribution parameterized by neural networks to approximate mutual information, enabling the derivation of lower bounds for effective model optimization. (3) An additive angular margin penalty is introduced into the speaker classification loss to promote discriminative embeddings and preserve identity utility. Experiments on public speech datasets demonstrate that SIBNet significantly outperforms state of the art methods in both biometric recognition accuracy and suppression of sensitive soft biometric attributes.

Privacy and security of biometric soft attributes in facial recognition images are gradually gaining people’s attention. The most cutting-edge solutions currently available can protect biometric soft attributes while preserving identity utility and enabling the reconstruction of original images from the protected ones. However, these solutions have shortcomings in that: 1) they do not allow users to choose the attributes to be protected freely and tend to make only minor modifications to the images for easier reconstruction; 2) during reconstruction, they can only restore common features of the attributes (such as adding eye makeup and lipstick to female images) and cannot reproduce the specific expressions of the attributes. In response to these limitations, we design a solution called NouMenn, which allows for the free selection of protected attributes and enables precise reconstruction. NouMenn utilizes style codes from the original images to guide image reconstruction, enabling a more accurate reproduction of specific attribute expressions in the images. Additionally, NouMenn permits more extensive modifications to the images during privacy protection, as there is sufficient information to support image reconstruction. To further enhance the security performance of the solution and avoid direct storage of style codes, we use Syndrome-Trellis Code to embed style code into the protected images. Users only need the weight matrix to stegoanalyze the stego protected images to get the style code. Extensive experiments demonstrate that NouMenn can effectively protect the privacy of facial soft attributes and achieve precise reconstruction, with both the resulting protected and reconstructed images retaining identity utility.

Abstract Traffic flow prediction plays a crucial role in intelligent transportation systems as it enables effective control and management of urban traffic. However, existing methods that based on Graph Convolutional Networks (GCNs) primarily utilize local neighborhood information for message passing, resulting in limited perception of global structures. Additionally, it is also a challenge to extract spatial-temporal similarity features due to the constraints of graph structures. To address these issues, we propose a novel traffic flow prediction model based on Dynamic Spatial-Temporal Similarity Pyramid Network (DSTSPYN). Our model employs a spatial-temporal pyramid architecture, which dynamically adjusts the weights of central, edge, and global spatial-temporal features using an enhanced attention mechanism. Furthermore, it captures dynamic temporal dependencies at different scales through pyramid gated convolution. Meanwhile, the spatial similarity features of different time steps can be extracted through the spatial-temporal global similarity (STGS) module. We evaluate our model on four public transportation datasets and demonstrate that the DSTSPYN model outperforms several baseline methods in terms of prediction accuracy. It effectively captures the dynamic spatial-temporal correlations of the road network and edge node features, making it well-suited for long-term traffic flow prediction.

Abstract In recent years, the proliferation of facial image collection systems coupled with significant advancements in machine learning-driven image analysis techniques has posed formidable challenges to protecting individuals’ privacy information, raising concerns about the security of such sensitive data. The current state-of-the-art technology is adept at extracting an array of intimate personal privacy details, encompassing gender, race, and potentially more, from a solitary facial image, underscoring the intricacies and implications of data privacy. Therefore, there is an urgent need for research on models that can protect the privacy of facial images. To tackle this issue, we proposed a multi-attribute privacy-preserving computational model based on generative adversarial networks (MAP-GAN) to protect sensitive facial privacy attribute information at the image level. For MAP-GAN, we meticulously design a privacy preservation loss function and introduce an attribute probability scoring mechanism to address the problem of binary attribute privacy protection flipping observed in previous models. Additionally, we incorporate an $$\textit{L}_1$$ L 1 information loss constraint to ensure the practical superiority of MAP-GAN by maintaining the information gap between privacy-protected images and original images. To enhance the quality of generated privacy-preserving images, we present a privacy-preserving image generator that utilizes residual structures and selective transmission units in MAP-GAN’s design. Experimental results demonstrate that MAP-GAN outperforms other models in terms of both multi-attribute privacy protection and utility.

Consumer electronic devices with voice assistants are becoming increasingly popular in modern intelligent home. Nevertheless, directly uploading unprocessed speech data, which may contain sensitive attributes, to a cloud server poses a significant risk to user privacy. To address this privacy issue, this paper proposes a privacy-enhancing model to protect speech emotions based on generative adversarial networks (PSEGAN). The model aims to prevent the inference of emotional attributes while maintaining the accuracy and utility of speech features. PSEGAN benefits from three modules: (1) A pre-trained speaker matcher imposes generative constraints on the model during the training phase to ensure that the generated speech retains the essential information needed for speaker recognition. (2) Attribute adversarial networks can generate perturbed speech that transforms emotional attributes while preserving the utility of the speech. (3) Gated Recurrent Networks (GRN) can handle the long-short term dependencies of speech signals. PSEGAN model solves the problem of utility loss in traditional speech privacy preservation methods based on generative adversarial networks (GAN). Experimental results show that on the RAVDESS dataset, PSEGAN reduces emotion recognition accuracy by 80.7%, while speaker recognition accuracy only decreases by 1.1%. These findings demonstrate that PSEGAN effectively mitigates the leakage of emotional attributes while maintaining high utility.

Soft-biometric privacy-enhancement methods are widely used in face recognition systems to prevent attackers from inferring soft-biometric attributes (e.g., gender, age, and race). However, existing methods typically focus on either representation-level or image-level privacy protection. In this paper, we propose a novel Face Conditional Privacy Funnel Network (Face-CPFNet), a dual-level privacy-enhancement framework with three key innovations. First, it introduces a dual-level privacy protection system where attackers receive reconstructed face images by combining the random guess of the sensitive attribute's potential label with the face representation provided by the user. Second, we propose the Face-CPF optimization problem, based on the Conditional Privacy Funnel (CPF) and an additional information leakage constraint. A deep variational approximation approach with parameterized deep neural networks is used to solve this problem and develop the Face-CPFNet model. Third, it enables supervised disentangled representation learning for reconstructing face images with variations in generative factors, and further enhances privacy protection by introducing an information leakage constraint when the sensitive attribute is a discrete binary random variable. Experimental results on benchmark datasets indicate that Face-CPFNet strikes a more effective balance between face verification accuracy and soft-biometric privacy compared with existing SBPE and CPF models.

Deep Neural Networks (DNNs) have been extensively employed for automatic face recognition, enabling the extraction of compact and discriminative representations from facial images. However, these representations typically encode a multitude of information ranging from individual identities to sensitive soft-biometric attributes such as gender, race, or age. This raises concerns regarding the privacy disclosure of soft-biometric as these attributes should be protected. To address this issue, we propose a novel Face Information Bottleneck Network (FIBNet), which is a representation-level privacy-enhancing framework based on the Information Bottleneck (IB) principle. The proposed FIBNet differs significantly from previous representation-level privacy-enhancing techniques in three key aspects. First, it generates a privacy-enhanced face representation, providing novel insights through an information-theoretic privacy framework. Second, we formulate the privacy protection of soft-biometric attributes as an IB optimization problem by striking a tradeoff between preserving a controlled amount of identity information within face representations and suppressing soft-biometric attribute information. Last, the proposed approach protects soft-biometric privacy from adversaries interested in specific sensitive attributes that are unknown to the biometric system designers or users. Detailed experimental results obtained on widely recognized facial recognition datasets demonstrate that the proposed FIBNet significantly outperforms the state-of-the-art methods in terms of both biometric performance for face verification and its soft-biometric attribute suppression efficiency. These notable results verify FIBNet as a novel and effective approach for ensuring representation-level soft-biometric privacy.

The applications of social Internet of Things (SIoT) with large numbers of intelligent devices provide a novel way for social behaviors. Intelligent devices share images according to the groups of their specified owners. However, sharing images may cause privacy disclosure when the images are illegally distributed without owners’ permission. To tackle this issue, combining blind watermark with additive secret sharing technique, we propose a lightweight and privacy-preserving image sharing (LPIS) scheme with illegal distributor detection in SIoT. Specifically, the query user’s authentication information is embedded in two shares of the transformed encrypted image by using discrete cosine transform (DCT) and additive secret sharing technique. The robustness against attacks, such as JPEG attack and the least significant bit planes (LSBs) replacement attacks, are improved by modifying 1/8 of coefficients of the transformed image. Moreover, we adopt two edge servers to provide image storage and authentication information embedding services for reducing the operational burden of clients. As a result, the identity of the illegal distributor can be confirmed by the watermark extraction of the suspicious image. Finally, we conduct security analysis and ample experiments. The results show that LPIS is secure and robust to prevent illegal distributors from modifying images and manipulating the embedded information before unlawful sharing.

With the widespread adoption of mobile devices, an increasingly number of personal data are stored in mobile devices that using flash memory as storage medium. Personal data privacy may also be leaked because of unauthorized access or resale of mobile devices. How to effectively protect users' data privacy and securely delete invalid data, which brings a great challenge to the data secure deletion in flash memory. In order to tackle these problems, we propose a novel data secure deletion scheme based on key derivation encryption algorithm for mobile devices. Firstly, we construct a node key tree based on flash hierarchical structure, and propose a key derivation encryption algorithm to generate data key to encrypt user data. Furthermore, we combine partial block erasure with partial key deletion method to delete both the ciphertext data and the partial key component after expired. The security analysis shows that the proposed scheme is able to implement data privacy protect and secure deletion of invalid data. Performance analysis and experimental results indicate that the proposed scheme is effective and efficient.

Mobile crowdsensing as a novel service schema of the Internet of Things (IoT) provides an innovative way to implement ubiquitous social sensing. How to establish an effective mechanism to improve the participation of sensing users and the authenticity of sensing data, protect the users’ data privacy, and prevent malicious users from providing false data are among the urgent problems in mobile crowdsensing services in IoT. These issues raise a gargantuan challenge hindering the further development of mobile crowdsensing. In order to tackle the above issues, in this paper, we propose a reliable hybrid incentive mechanism for enhancing crowdsensing participations by encouraging and stimulating sensing users with both reputation and service returns in mobile crowdsensing tasks. Moreover, we propose a privacy preserving data aggregation scheme, where the mediator and/or sensing users may not be fully trusted. In this scheme, differential privacy mechanism is utilized through allowing different sensing users to add noise data, then employing homomorphic encryption for protecting the sensing data, and finally uploading ciphertext to the mediator, who is able to obtain the collection of ciphertext of the sensing data without actual decryption. Even in the case of partial sensing data leakage, differential privacy mechanism can still ensure the security of the sensing user’s privacy. Finally, we introduce a novel secure multiparty auction mechanism based on the auction game theory and secure multiparty computation, which effectively solves the problem of prisoners’ dilemma incurred in the sensing data transaction between the service provider and mediator. Security analysis and performance evaluation demonstrate that the proposed scheme is secure and efficient.

Rapid development of cloud storage services produces a tremendous amount of user data outsourcing to cloud servers. Therefore, it is easy to generate data multi-replica, which is able to improve data availability and users' experience. However, when the management of data is poor, the sensitive information will be disclosed more easily. This may bring serious security and privacy challenges for both user's data and its multi-replica in cloud environment. In order to tackle the above issues, in this paper, we propose a multi-replica associated deleting scheme (MADS) in cloud environment. We first introduce a replica associated model to organize all of data replicas among different cloud servers. Furthermore, we propose the MADS scheme which is consists of data storage algorithm, replica generation algorithm, replica deletion and feedback algorithm. Finally, we employ Amazon S3 to implement MADS and the results indicate that the proposed scheme is available and effective.

To enhance security and robustness of watermarking algorithms, this paper proposes a robust 3D model watermarking scheme based on non-uniform discrete Fourier transform (NDFT). In order to resist attacks such as non-uniform scale, rotation and translation, the affine invariant norm is applied to build partitions of vertices, which can be carried out by NDFT. The paper is also proved that the bit-embedding does not change the mass center of 3D model. The original model is not required in watermark detection. Simulation results show that the proposed scheme has high robustness against these attacks(including noise attack).