Federated learning (FL) has achieved state-of-the-art performance in distributed learning tasks with privacy requirements. However, it has been discovered that FL is vulnerable to adversarial attacks. The typical gradient inversion attacks primarily focus on attempting to obtain the client’s private input in a white-box manner, where the adversary is assumed to be either the client or the server. However, if both the clients and the server are honest and fully trusted, is the FL secure? In this paper, we propose a novel method called External Gradient Inversion Attack (EGIA) in the grey-box settings. Specifically, we concentrate on the point that public-shared gradients in FL are always transmitted through the intermediary nodes, which has been widely ignored. On this basis, we demonstrate that an external adversary can reconstruct the private input using gradients even if both the clients and the server are honest and fully trusted. We also provide a comprehensive theoretical analysis of the black-box attack scenario in which the adversary has only the gradients. We perform extensive experiments on multiple real-world datasets to test the effectiveness of EGIA. The outcomes of our experiments validate that the EGIA method is highly effective.

Most of the image encryption schemes based on chaos have so far employed symmetric key cryptography, which leads to a situation where the key cannot be transmitted in public channels, thus limiting their extended application. Based on the elliptic curve cryptography (ECC), we proposed a public key image encryption method where the hash value derived from the plain image was encrypted by ECC. Furthermore, during image permutation, a novel algorithm based on different-sized block was proposed. The plain image was firstly divided into five planes according to the amount of information contained in different bits: the combination of the low 4 bits, and other four planes of high 4 bits respectively. Second, for different planes, the corresponding method of block partition was followed by the rule that the higher the bit plane, the smaller the size of the partitioned block as a basic unit for permutation. In the diffusion phase, the used hyperchaotic sequences in permutation were applied to improve the efficiency. Lots of experimental simulations and cryptanalyses were implemented in which the NPCR and UACI are 99.6124% and 33.4600% respectively, which all suggested that it can effectively resist statistical analysis attacks and chosen plaintext attacks.

The redactable blockchain has emerged as a promising technique in mobile crowdsensing, allowing users to break immutability in a controlled manner selectively. Unfortunately, current fine-grained redactable blockchains suffer two significant limitations in terms of security and functionality, which severely impede their application in mobile crowdsensing. For security, the transparency of the blockchain allows anyone to access both the data and policy, which consequently results in a breach of user privacy. Regarding functionality, current solutions cannot support error tolerance during policy matching, thereby limiting their applicability in various situations, such as fingerprint-based and face-based identification scenarios. This paper presents a privacy-preserving fine-grained redactable blockchain with policy fuzzy matching, named PRBFM. PRBFM supports fuzzy policy matching and partitions users’ privileges without compromising user privacy. The idea of PRBFM is to leverage threshold linear secret sharing based on the Lagrange interpolation theorem to distribute the decryption keys and chameleon hash trapdoors. Additionally, we have incorporated a privacy-preserving policy matching delegation mechanism into PRBFM to minimize user overhead. Our security analysis demonstrates that PRBFM can defend against the chosen-ciphertext attack. Moreover, experiments conducted on the FISCO blockchain platform show that PRBFM is at least 7.8 times faster than existing state-of-the-art solutions.

To boost the encoder stealing attack under the perturbation-based defense that hinders the attack performance, we propose a boosting encoder stealing attack with perturbation recovery named BESA. It aims to overcome perturbation-based defenses. The core of BESA consists of two modules: perturbation detection and perturbation recovery, which can be combined with canonical encoder stealing attacks. The perturbation detection module utilizes the feature vectors obtained from the target encoder to infer the defense mechanism employed by the service provider. Once the defense mechanism is detected, the perturbation recovery module leverages the well-designed generative model to restore a clean feature vector from the perturbed one. Through extensive evaluations based on various datasets, we demonstrate that BESA significantly enhances the surrogate encoder accuracy of existing encoder stealing attacks by up to 24.63\% when facing state-of-the-art defenses and combinations of multiple defenses.