Simin Yu

Qianxue Wang, Simin Yu

Plaintext non-delayed chaotic cipher (PNDCC) means that in the diffusion equation, plaintext has no delay terms while ciphertext has a feedback term. In existing literature, chaotic cipher diffusions invariably take this form. Since its introduction, PNDCC has attracted attention but also doubts. Designers of chaotic ciphers usually claim PNDCC security by statistical tests, while rigorous cryptographic proofs are absent. Thus, it is necessary to re-examine its design rationale and empirical security. To address this issue, we present a typical example of a three-stage permutation-diffusion-permutation PNDCC, which contains multiple security vulnerabilities. Although all of its statistical indicators show good performance, we are able to break it using four different attacks. The first is a differential attack based on homogeneous operations; the second is an S-PTC attack; the third is a novel impulse-step-based differential attack (ISBDA), proposed in this paper, and the fourth is a novel chain attack, also introduced here. These results demonstrate that the fulfilment of statistical criteria is not a sufficient condition for the security of PNDCC. Then, based on a mathematical model of multi-stage PNDCC, we show that the proposed chain attack can successfully break a class of multi-stage PNDCCs. The key technique of the chain attack depends on how to reveal all permutations. To address this key problem, we summarize the chaining rules and show that, from the attacker's perspective, if the same decryption chain can be reconstructed then all permutations can be deciphered. To that end, the entire diffusion process can be broken by solving a system of simultaneous equations. Finally, as a secure improvement, we propose a new scheme termed plaintext-delayed chaotic cipher (PDCC) that can resist various cryptanalytic attacks.

Hao Fang, Xiaohang Sui, Hongyao Yu et al.

Diffusion models (DMs) have recently demonstrated remarkable generation capability. However, their training generally requires huge computational resources and large-scale datasets. To solve these, recent studies empower DMs with the advanced Retrieval-Augmented Generation (RAG) technique and propose retrieval-augmented diffusion models (RDMs). By incorporating rich knowledge from an auxiliary database, RAG enhances diffusion models' generation and generalization ability while significantly reducing model parameters. Despite the great success, RAG may introduce novel security issues that warrant further investigation. In this paper, we reveal that the RDM is susceptible to backdoor attacks by proposing a multimodal contrastive attack approach named BadRDM. Our framework fully considers RAG's characteristics and is devised to manipulate the retrieved items for given text triggers, thereby further controlling the generated contents. Specifically, we first insert a tiny portion of images into the retrieval database as target toxicity surrogates. Subsequently, a malicious variant of contrastive learning is adopted to inject backdoors into the retriever, which builds shortcuts from triggers to the toxicity surrogates. Furthermore, we enhance the attacks through novel entropy-based selection and generative augmentation strategies that can derive better toxicity surrogates. Extensive experiments on two mainstream tasks demonstrate the proposed BadRDM achieves outstanding attack effects while preserving the model's benign utility.

Sijin Yu, Zijiao Chen, Wenxuan Wu et al.

Reconstructing visual stimuli from human brain activity (e.g., fMRI) bridges neuroscience and computer vision by decoding neural representations. However, existing methods often overlook critical brain structure-function relationships, flattening spatial information and neglecting individual anatomical variations. To address these issues, we propose (1) a novel sphere tokenizer that explicitly models fMRI signals as spatially coherent 2D spherical data on the cortical surface; (2) integration of structural MRI (sMRI) data, enabling personalized encoding of individual anatomical variations; and (3) a positive-sample mixup strategy for efficiently leveraging multiple fMRI scans associated with the same visual stimulus. Collectively, these innovations enhance reconstruction accuracy, biological interpretability, and generalizability across individuals. Experiments demonstrate superior reconstruction performance compared to SOTA methods, highlighting the effectiveness and interpretability of our biologically informed approach.

Zhuosheng Lin, Christophe Guyeux, Qianxue Wang et al.

To guarantee the integrity and security of data transmitted through the Internet, hash functions are fundamental tools. But recent researches have shown that security flaws exist in the most widely used hash functions. So a new way to improve their security performance is urgently demanded. In this article, we propose new hash functions based on chaotic iterations, which have chaotic properties as defined by Devaney. The corresponding diffusion and confusion analyzes are provided and a comparative study between the proposed hash functions is carried out, to make their use more applicable in any security context.

Zhuosheng Lin, Christophe Guyeux, Simin Yu et al.

Investigating how to construct a secure hash algorithm needs in-depth study, as various existing hash functions like the MD5 algorithm have recently exposed their security flaws. At the same time, hash function based on chaotic theory has become an emerging research in the field of nonlinear information security. As an extension of our previous research works, a new chaotic iterations keyed hash function is proposed in this article. Chaotic iterations are used both to construct strategies with pseudorandom number generator and to calculate new hash values using classical hash functions. It is shown that, by doing so, it is possible to apply a kind of post-treatment on existing hash algorithms, which preserves their security properties while adding Devaney's chaos. Security performance analysis of such a post-treatment are finally provided.

Eric Yong Xie, Chengqing Li, Simin Yu et al.

Utilizing complex dynamics of chaotic maps and systems in encryption was studied comprehensively in the past two and a half decades. In 1989, Fridrich's chaotic image encryption scheme was designed by iterating chaotic position permutation and value substitution some rounds, which received intensive attention in the field of chaos-based cryptography. In 2010, Solak \textit{et al.} proposed a chosen-ciphertext attack on the Fridrich's scheme utilizing influence network between cipher-pixels and the corresponding plain-pixels. Based on their creative work, this paper scrutinized some properties of Fridrich's scheme with concise mathematical language. Then, some minor defects of the real performance of Solak's attack method were given. The work provides some bases for further optimizing attack on the Fridrich's scheme and its variants.

Qianxue Wang, Simin Yu, Christophe Guyeux et al.

In this paper, a new approach for constructing integer domain chaotic systems (IDCS) is proposed, and its chaotic behavior is mathematically proven according to the Devaney's definition of chaos. Furthermore, an analog-digital hybrid circuit is also developed for realizing the designed basic IDCS. In the IDCS circuit design, chaos generation strategy is realized through a sample-hold circuit and a decoder circuit so as to convert the uniform noise signal into a random sequence, which plays a key role in circuit implementation. The experimental observations further validate the proposed systematic methodology for the first time.