The widespread adoption of Transformer models raises critical privacy concerns as users must expose sensitive inputs to service providers during inference. While existing secure Transformer frameworks have addressed this issue to some extent, most rely on interactive protocols with prohibitive communication overhead, limiting practicality in bandwidth-constrained scenarios. This paper introduces Zephyr, a secure and non-interactive two-party inference framework for Transformers that overcomes these limitations. First, Zephyr proposes two novel SIMD ciphertext decompression techniques, shifting from serial to parallel processing to accelerate decompression by 1.2× while preserving accuracy. Second, Zephyr optimizes the deployment strategy of bootstrapping operations (which refresh encrypted data noise) during computation. This allows using smaller encryption parameters while achieving 1.1× faster bootstrapping than NEXUS. Evaluated on BERT-base under challenging 100Mbps/80ms conditions, Zephyr demonstrates superior performance - 24.3× faster than Iron(NeurIPS22), 2.1× faster than BOLT, and 11% faster than NEXUS while reducing communication costs by 95% versus BOLT(Oakland24) and 32% versus NEXUS(NDSS25), making it particularly effective for bandwidth-constrained environments while maintaining security against semi-honest adversaries.

Federated learning (FL) allows for collaborative training without centralizing data, but concerns regarding model privacy leakage, intellectual property theft and poisoning attacks have hindered its development. To mitigate such risks, this paper proposes embedded message authentication code technology (EMAC) to integrate encryption, digital signatures, and watermark functions for model security. In EMAC, the authentication data is embedded into the model ciphertext using reversible data hiding after encryption. The marked ciphertext supports data extraction for subsequent authentication and lossless decryption for testing and training simultaneously. Based on EMAC, a novel FL with security authentication and traceability of poisoning (FL-SATP) is proposed, which integrates privacy protection, identity authentication and poisoning traceability into FL. The poisoner tracing is designed to detect and identify poisoners retrospectively based on the practical performance of trained or aggregated models, thus removing the malicious users' model and deterring poisoning behaviors. Theoretical analysis and experimental results demonstrate that FL-SATP could ensure the confidentiality of the model content, the availability of model function, and that when more than half of the users are benign, the proposed method can accurately and efficiently pinpoint all malicious poisoners in FL.

Privacy-preserving multidimensional data aggregation aggregates the data of all different users into a single value, preventing the leakage of personal data while ensuring its availability. However, most current multidimensional data aggregation schemes only consider sum operations, i.e., the message of each dimension in the aggregation result is the sum of the corresponding dimensional messages of all individual message vectors. We propose a multidimensional data aggregation scheme with fine-grained linear homomorphism, called MDA-FLH. Firstly, we construct a fine-grained linear homomorphic encryption scheme which can assign different weights to each dimension of user’s data and maintain the linear homomorphic property in each dimension. We combine the Chinese remainder theorem and Paillier encryption to encode the multidimensional data with CRT and assign different weights to each dimension of user’s data in the Paillier ciphertext. Secondly, our scheme has the property of fault tolerance. In conjunction with extended Shamir’s threshold secret-sharing scheme, a security-enhanced and fault-tolerant data aggregation method has been designed so that it is resistant to internal attacks such as the control center (CC) access to individual private data if given the corresponding ciphertext. Finally, two practical schemes are designed based on MDA-FLH: fine-grained electricity price statistics scheme and multistep electricity price statistics scheme. Security analysis shows that our scheme can achieve privacy, confidentiality, integrity, and source authentication. Performance analysis shows that our scheme is efficient, especially that SMs are computationally economic, which makes our scheme more suitable for resource-constrained SM. Also, our scheme can provide linear homomorphism operations on each dimension, which further expands its applications.

Privacy protection of biometrics-based on cloud computing is attracting increasing attention. In 2018, Zhuet al.proposed an efficient and privacy-preserving online fingerprint authentication scheme for data outsourcing e-Finga. Under the premise of ensuring user's fingerprint data privacy and message security authentication, the e-Finga scheme can provide accurate and efficient fingerprint identity authentication services. However, our analysis shows that the temporary fingerprint in this scheme uses the deterministic encryption algorithm, which has the risk of leaking the user's fingerprint characteristics. Therefore, we propose a temporary fingerprint attack method for the e-Finga scheme. Experiments demonstrate that an adversary can analyze specific secret parameters and fingerprint features when eavesdropping on a user's temporary fingerprint ciphertext. To counter the temporary fingerprint attack, we propose a secure e-fingerprint scheme– Secure e-finger that uses the learning with errors samples, which has the homomorphic addition property, to encrypt user's temporary fingerprints. Experiments show that the secure e-finger scheme can resist the temporary fingerprint attack. Compared with the unprotected e-Finga scheme, the client running time is increased by about 6% percent, the communication cost on the user side only increased by 0.3125% percent. As a result, our solution can realize secure online fingerprint authentication without losing efficiency. Single user authentication is likely to cause the problem of excessive authority. Based on the Secure e-finger scheme, we propose a threshold scheme based on biological characteristics.

Biometric technologies are being prominently used everywhere. However, the leakage of biometric information can pose a serious security risk, making the protection of biometric templates particularly important and receiving more attention. Rathgeb et al. first proposed the cancelable biometric technology based on Bloom filters in 2013, which has been applied to protect different biometric templates. Bloom filter-based biometrics offer the advantages of alignment-free, fast recognition and high accuracy. An ideal biometric system should also be irreversibility and unlinkability. In this paper, firstly, we propose a reverse reconstruction attack. Through the reverse reconstruction of Bloom filters, we find that the reconstructed biometric data and the original biometric data have some strong statistical correlation, which proves that the scheme has the linkability defect. Experiments show that for the original Bloom filter-based biometric template protection scheme, we can judge whether two different biometric templates belong to the same user with a success probability of 71.0%. Secondly, to remedy above defect, we construct a structure-preserving encryption scheme, i.e., the feature template encrypted with it maintains the structure and length of the original template, making it impossible for an attacker to reconstruct meaningful data from Bloom filter. Finally, an improved biometric template protection scheme based on Bloom filters is proposed by introducing the proposed encryption. Attack experiment shows that the improved scheme can effectively resist the reverse reconstruction attack, with the success probability of attacking the unlinkability of the improved scheme being 50.0%, which is the same as the probability of random guessing. Performance evaluation shows that the proposed scheme maintains the biometric performance of the original system and the unprotected system.

Abstract Multi‐key fully homomorphic encryption (MKFHE) allows computations on ciphertexts encrypted by different users, which can be applied to implement secure multi‐party computing (MPC). The current NTRU‐based MKFHE has the following two drawbacks: One is that the relinearisation process during homomorphic evaluation is so complicated that the corresponding computation time is costly. The other is that a class of subfield attacks are proposed and affects the security of NTRU schemes over power‐of‐2 cyclotomic rings for large moduli q, especially for the NTRU‐based fully homomorphic encryption (FHE) schemes. In this work, an efficient MKFHE scheme is proposed over prime cyclotomic rings with fewer relinearisations, which seems a good choice because of its potential to resist a subfield attack. More specifically, the time of the relinearisation process is reduced by half in homomorphic evaluations by separating the homomorphic multiplication and the relinearisation process (implementing two homomorphic multiplication operations together before relinearisation), while in current NTRU‐type MKFHE schemes, these two processes are usually performed together. The error bound of the basic function components is re‐analysed over prime cyclotomic rings in the average case, which can be used in the error analysis of our scheme. We construct an efficient NTRU‐based single‐key FHE scheme and an efficient MKFHE scheme over prime cyclotomic rings through relinearisation and modulus‐switching techniques. The MKFHE scheme proposed has the on‐the‐fly property and has a tight ciphertext size compared with the GSW‐type and BGV‐type MKFHE schemes. An experiment shows that the homomorphic evaluation of the optimised single‐key FHE scheme proposed is 1.9 times faster than an efficient NTRU‐type MKFHE DHS16 proposed at DCC 2016.