Given the immutability of biometric data, it is imperative to develop a biometric template protection method that guarantees the complete non-disclosure of any original biometric information while ensuring high recognition performance. Two mainstream approaches in biometric template protection—cancelable biometrics and biometric cryptosystems—have been widely adopted; however, the protected templates produced by these methods still contain some of the original biometric data, which can lead to privacy leakage. To address these challenges, we propose a novel framework named BioDeepHash that integrates deep hashing with cryptographic hash functions. In our approach, a deep hashing model generates consistent templates for similar biometric data from the same user, thereby eliminating intra-class variations. An application-specific XOR string is then used to achieve revocability, and finally these consistent templates are processed by cryptographic hash functions to produce protected templates that meet strict security standards. Our experimental results show that, compared with existing methods, BioDeepHash increases the average Genuine Acceptance Rate by 10.12% on the iris dataset and by 3.12% on the facial dataset, while achieving an extremely low False Acceptance Rate—0% for the iris dataset and only 0.0002% for the facial dataset.

With the development of genome sequencing technology, genomic data have been widely collected and used in real-world scenarios, such as genomic medicine and similar patient query (SPQ) cases, based on similarity comparisons of genomic sequences. However, genomic data are unique for every person and contain a large amount of sensitive information involving personal privacy. Therefore, determining how to protect privacy while utilizing genomic data has become a key issue. In this paper, we mainly investigate the privacy protection of SPQs, and we propose five algorithms (NDB-ED, NDB-Band, NDB-Block, NDB-SIS and NDB-SDS) based on a promising technique called negative representation of information (NRI). The proposed algorithms use five kinds of similarity comparison approaches widely applied in SPQs and convert all genomic sequences into negative databases (NDBs, among the most important NRI forms) for privacy protection. When performing an SPQ, NDB-ED approximates the edit distance between the sketches (the statistics of NDBs) of two genomic sequences to evaluate the dissimilarity. Banded edit distance and block-wise edit distance are two effective approximate edit distance algorithms, which can greatly reduce the time complexity. NDB-Band and NDB-Block are used to estimate the banded edit distance and the block-wise edit distance between the sketches of NDB pairs, respectively, to further improve query efficiency and reduce communication overhead. Besides, private genome set intersection size (SIS) and set difference size (SDS) can also be used instead of edit distance to evaluate the dissimilarity between genomic pairs during SPQs. NDB-SIS and NDB-SDS estimate the SIS and SDS between the sketches, for similarity comparison. The experimental results demonstrate that the proposed algorithms can achieve promising results in terms of accuracy and efficiency (Our best algorithm improves accuracy by at least 10% and query time is at least 700 times faster than existing algorithms during SPQs).

In recent years, deep learning has been applied in a wide variety of domains and gains outstanding success. In order to achieve high accuracy, a large amount of training data and high-performance hardware are necessary for deep learning. In real-world applications, many deep learning developers usually rent cloud GPU servers to train or deploy their models. Since training data may contain sensitive information, training models on cloud servers will cause severe privacy leakage problem. To solve this problem, we propose a privacy-preserving deep learning model based on matrix transformation. Specifically, we transform original data by adding or multiplying a random matrix. The obtained data is significantly different from the origin and it is hard to recover original data, so it can protect the privacy in original data. Experimental results demonstrate that the models trained with processed data can achieve high accuracy.

The application of iris biometric data has become prevalent across various domains, encompassing access control, identity verification, and criminal investigations. Consequently, there is a pressing need to develop effective methods for safeguarding the privacy of iris data. While numerous methods for iris data protection have been proposed, the majority of them fall short of meeting the ISO/IEC 24745 standards about irreversibility, revocability, and unlinkability. In this paper, we introduce a novel iris data protection method called TNCB, which is based on a transformation network. The TNCB involves performing a block-wise permutation of the original iris images using application-specific parameters, followed by pixel-by-pixel modulo and inversion fusion operations. The resulting images are subsequently employed for pre-training a recognition network that will be used to recognize protected images. Afterwards, a transformation network is introduced to achieve a further non-invertible transformation. Our security analysis demonstrates that the TNCB could fulfill the three major security requirements. To validate its effectiveness, we conducted a series of attack and performance experiments on the CASIA-Iris-Lamp and CASIA-Iris-Thousand datasets. Experimental results substantiated the robustness of TNCB in maintaining recognition performance while safeguarding the privacy of iris data. Furthermore, experimental results also highlight that our scheme could effectively support iris recognition in both open-set and close-set modes.

In recent years, the rapid development of deep learning is attributed to the large amount of labeled data brought by the digital age. When there is no labeled data available in some application scenarios, domain adaptation can be used to transfer knowledge from the source domain with labeled data to the target domain without labeled data. In the process of domain adaptation, the target client requires direct access to the source data or model, which would lead to the risk of privacy leakage, e.g., Membership Inference Attacks (MIA). Attackers can collect the prediction vector of the model through black-box access to the source model, and then infer an individual’s membership in the source training dataset. To deal with this privacy issue, we propose a GAN-based Privacy-Preserving Unsupervised Domain Adaptation Framework. Specifically, the target client learns a conditional generator, sends the intermediate results perturbed by differential privacy to the source client, and the source client uses the source model to provide guidance for the generator so that the generator can generate the data corresponding to the input label that is the same as the data distribution in the target domain. We evaluate the performance of our proposed method on digital dataset and office-31dataset, which are popular domain adaptation benchmark datasets, and verify the security by the accuracy and F1-score of Membership Inference Attacks.