Federated Learning Backdoor Attack Based on Frequency Domain Injection.

Federated learning (FL) is a distributed machine learning framework that enables scattered participants to collaboratively train machine learning models without revealing information to other participants. Due to its distributed nature, FL is susceptible to being manipulated by malicious clients. These malicious clients can launch backdoor attacks by contaminating local data or tampering with local model gradients, thereby damaging the global model. However, existing backdoor attacks in distributed scenarios have several vulnerabilities. For example, (1) the triggers in distributed backdoor attacks are mostly visible and easily perceivable by humans; (2) these triggers are mostly applied in the spatial domain, inevitably corrupting the semantic information of the contaminated pixels. To address these issues, this paper introduces a frequency-domain injection-based backdoor attack in FL. Specifically, by performing a Fourier transform, the trigger and the clean image are linearly mixed in the frequency domain, injecting the low-frequency information of the trigger into the clean image while preserving its semantic information. Experiments on multiple image classification datasets demonstrate that the attack method proposed in this paper is stealthier and more effective in FL scenarios compared to existing attack methods.

Secure Ring Signature Scheme for Privacy-Preserving Blockchain.

Blockchain integrates peer-to-peer networks, distributed consensus, smart contracts, cryptography, etc. It has the unique advantages of weak centralization, anti-tampering, traceability, openness, transparency, etc., and is widely used in various fields, e.g., finance and healthcare. However, due to its open and transparent nature, attackers can analyze the ledger information through clustering techniques to correlate the identities between anonymous and real users in the blockchain system, posing a serious risk of privacy leakage. The ring signature is one of the digital signatures that achieves the unconditional anonymity of the signer. Therefore, by leveraging Distributed Key Generation (DKG) and Elliptic Curve Cryptography (ECC), a blockchain-enabled secure ring signature scheme is proposed. Under the same security parameters, the signature constructed on ECC has higher security in comparison to the schemes using bilinear pairing. In addition, the system master key is generated by using the distributed key agreement, which avoids the traditional method of relying on a trusted third authorizer (TA) to distribute the key and prevents the key leakage when the TA is not authentic or suffers from malicious attacks. Moreover, the performance analysis showed the feasibility of the proposed scheme while the security was ensured.

Image Adversarial Example Generation Method Based on Adaptive Parameter Adjustable Differential Evolution.

Adversarial example generation techniques for neural network models have exploded in recent years. In the adversarial attack scheme for image recognition models, it is challenging to achieve a high attack success rate with very few pixel modifications. To address this issue, this paper proposes an adversarial example generation method based on adaptive parameter adjustable differential evolution. The method realizes the dynamic adjustment of the algorithm performance by adjusting the control parameters and operation strategies of the adaptive differential evolution algorithm, while searching for the optimal perturbation. Finally, the method generates adversarial examples with a high success rate, modifying just a very few pixels. The attack effectiveness of the method is confirmed in CIFAR10 and MNIST datasets. The experimental results show that our method has a greater attack success rate than the One Pixel Attack based on the conventional differential evolution. In addition, it requires significantly less perturbation to be successful compared to global or local perturbation attacks, and is more resistant to perception and detection.

Wasserstein Distance-Based Deep Leakage from Gradients.

Federated learning protects the privacy information in the data set by sharing the average gradient. However, "Deep Leakage from Gradient" (DLG) algorithm as a gradient-based feature reconstruction attack can recover privacy training data using gradients shared in federated learning, resulting in private information leakage. However, the algorithm has the disadvantages of slow model convergence and poor inverse generated images accuracy. To address these issues, a Wasserstein distance-based DLG method is proposed, named WDLG. The WDLG method uses Wasserstein distance as the training loss function achieved to improve the inverse image quality and the model convergence. The hard-to-calculate Wasserstein distance is converted to be calculated iteratively using the Lipschit condition and Kantorovich-Rubinstein duality. Theoretical analysis proves the differentiability and continuity of Wasserstein distance. Finally, experiment results show that the WDLG algorithm is superior to DLG in training speed and inversion image quality. At the same time, we prove through the experiments that differential privacy can be used for disturbance protection, which provides some ideas for the development of a deep learning framework to protect privacy.

Stable and Fast Deep Mutual Information Maximization Based on Wasserstein Distance.

Deep learning is one of the most exciting and promising techniques in the field of artificial intelligence (AI), which drives AI applications to be more intelligent and comprehensive. However, existing deep learning techniques usually require a large amount of expensive labeled data, which limit the application and development of deep learning techniques, and thus it is imperative to study unsupervised machine learning. The learning of deep representations by mutual information estimation and maximization (Deep InfoMax or DIM) method has achieved unprecedented results in the field of unsupervised learning. However, in the DIM method, to restrict the encoder to learn more normalized feature representations, an adversarial network learning method is used to make the encoder output consistent with a priori positively distributed data. As we know, the model training of the adversarial network learning method is difficult to converge, because there is a logarithmic function in the loss function of the cross-entropy measure, and the gradient of the model parameters is susceptible to the "gradient explosion" or "gradient disappearance" phenomena, which makes the training of the DIM method extremely unstable. In this regard, we propose a Wasserstein distance-based DIM method to solve the stability problem of model training, and our method is called the WDIM. Subsequently, the training stability of the WDIM method and the classification ability of unsupervised learning are verified on the CIFAR10, CIFAR100, and STL10 datasets. The experiments show that our proposed WDIM method is more stable to parameter updates, has faster model convergence, and at the same time, has almost the same accuracy as the DIM method on the classification task of unsupervised learning. Finally, we also propose a reflection of future research for the WDIM method, aiming to provide a research idea and direction for solving the image classification task with unsupervised learning.

Attribute-based multi-user collaborative searchable encryption in COVID-19.

With the outbreak of COVID-19, the government has been forced to collect a large amount of detailed information about patients in order to effectively curb the epidemic of the disease, including private data of patients. Searchable encryption is an essential technology for ciphertext retrieval in cloud computing environments, and many searchable encryption schemes are based on attributes to control user's search permissions to protect their data privacy. The existing attribute-based searchable encryption (ABSE) scheme can only implement the situation where the search permission of one person meets the search policy and does not support users to obtain the search permission through collaboration. In this paper, we proposed a new attribute-based collaborative searchable encryption scheme in multi-user setting (ABCSE-MU), which takes the access tree as the access policy and introduces the translation nodes to implement collaborative search. The cooperation can only be reached on the translation node and the flexibility of search permission is achieved on the premise of data security. ABCSE-MU scheme solves the problem that a single user has insufficient search permissions but still needs to search, making the user's access policy more flexible. We use random blinding to ensure the confidentiality and security of the secret key, further prove that our scheme is secure under the Decisional Bilinear Diffie-Hellman (DBDH) assumption. Security analysis further shows that the scheme can ensure the confidentiality of data under chosen-keyword attacks and resist collusion attacks.

GDP vs. LDP: A Survey from the Perspective of Information-Theoretic Channel.

The existing work has conducted in-depth research and analysis on global differential privacy (GDP) and local differential privacy (LDP) based on information theory. However, the data privacy preserving community does not systematically review and analyze GDP and LDP based on the information-theoretic channel model. To this end, we systematically reviewed GDP and LDP from the perspective of the information-theoretic channel in this survey. First, we presented the privacy threat model under information-theoretic channel. Second, we described and compared the information-theoretic channel models of GDP and LDP. Third, we summarized and analyzed definitions, privacy-utility metrics, properties, and mechanisms of GDP and LDP under their channel models. Finally, we discussed the open problems of GDP and LDP based on different types of information-theoretic channel models according to the above systematic review. Our main contribution provides a systematic survey of channel models, definitions, privacy-utility metrics, properties, and mechanisms for GDP and LDP from the perspective of information-theoretic channel and surveys the differential privacy synthetic data generation application using generative adversarial network and federated learning, respectively. Our work is helpful for systematically understanding the privacy threat model, definitions, privacy-utility metrics, properties, and mechanisms of GDP and LDP from the perspective of information-theoretic channel and promotes in-depth research and analysis of GDP and LDP based on different types of information-theoretic channel models.

An Efficient Plaintext-Related Chaotic Image Encryption Scheme Based on Compressive Sensing.

With the development of mobile communication network, especially 5G today and 6G in the future, the security and privacy of digital images are important in network applications. Meanwhile, high resolution images will take up a lot of bandwidth and storage space in the cloud applications. Facing the demands, an efficient and secure plaintext-related chaotic image encryption scheme is proposed based on compressive sensing for achieving the compression and encryption simultaneously. In the proposed scheme, the internal keys for controlling the whole process of compression and encryption is first generated by plain image and initial key. Subsequently, discrete wavelets transform is used in order to convert the plain image to the coefficient matrix. After that, the permutation processing, which is controlled by the two-dimensional Sine improved Logistic iterative chaotic map (2D-SLIM), was done on the coefficient matrix in order to make the matrix energy dispersive. Furthermore, a plaintext related compressive sensing has been done utilizing a measurement matrix generated by 2D-SLIM. In order to make the cipher image lower correlation and distribute uniform, measurement results quantified the 0∼255 and the permutation and diffusion operation is done under the controlling by two-dimensional Logistic-Sine-coupling map (2D-LSCM). Finally, some common compression and security performance analysis methods are used to test our scheme. The test and comparison results shown in our proposed scheme have both excellent security and compression performance when compared with other recent works, thus ensuring the digital image application in the network.

Wasserstein Generative Adversarial Networks Based Differential Privacy Metaverse Data Sharing.

Although differential privacy metaverse data sharing can avoid privacy leakage of sensitive data, randomly perturbing local metaverse data will lead to an imbalance between utility and privacy. Therefore, this work proposed models and algorithms of differential privacy metaverse data sharing using Wasserstein generative adversarial networks (WGAN). Firstly, this study constructed the mathematical model of differential privacy metaverse data sharing by introducing appropriate regularization term related to generated data's discriminant probability into WGAN. Secondly, we established basic model and algorithm for differential privacy metaverse data sharing using WGAN based on the constructed mathematical model, and theoretically analyzed basic algorithm. Thirdly, we established federated model and algorithm for differential privacy metaverse data sharing using WGAN by serialized training based on basic model, and theoretically analyzed federated algorithm. Finally, based on utility and privacy metrics, we conducted a comparative analysis for the basic algorithm of differential privacy metaverse data sharing using WGAN, and experimental results validate theoretical results, which show that algorithms of differential privacy metaverse data sharing using WGAN maintaining equilibrium between privacy and utility.

Low distortion reversible database watermarking based on hybrid intelligent algorithm.

In many fields, such as medicine and the computer industry, databases are vital in the process of information sharing. However, databases face the risk of being stolen or misused, leading to security threats such as copyright disputes and privacy breaches. Reversible watermarking techniques ensure the ownership of shared relational databases, protect the rights of data owners and enable the recovery of original data. However, most of the methods modify the original data to a large extent and cannot achieve a good balance between protection against malicious attacks and data recovery. In this paper, we proposed a robust and reversible database watermarking technique using a hash function to group digital relational databases, setting the data distortion and watermarking capacity of the band weight function, adjusting the weight of the function to determine the watermarking capacity and the level of data distortion, using firefly algorithms (FA) and simulated annealing algorithms (SA) to improve the efficiency of the search for the location of the watermark embedded and, finally, using the differential expansion of the way to embed the watermark. The experimental results prove that the method maintains the data quality and has good robustness against malicious attacks.