Designing an unsupervised image denoising approach in practical applications is a challenging task due to the complicated data acquisition process. In the realworld case, the noise distribution is so complex that the simplified additive white Gaussian (AWGN) assumption rarely holds, which significantly deteriorates the Gaussian denoisers’ performance. To address this problem, we apply a deep neural network that maps the noisy image into a latent space in which the AWGN assumption holds, and thus any existing Gaussian denoiser is applicable. More specifically, the proposed neural network consists of the encoder-decoder structure and approximates the likelihood term in the Bayesian framework. Together with a Gaussian denoiser, the neural network can be trained with the input image itself and does not require any pre-training in other datasets. Extensive experiments on real-world noisy image datasets have shown that the combination of neural networks and Gaussian denoisers improves the performance of the original Gaussian denoisers by a large margin. In particular, the neural network+BM3D method significantly outperforms other unsupervised denoising approaches and is competitive with supervised networks such as DnCNN, FFDNet, and CBDNet.

We state a conjecture for the formulas of the depth 4 low-weight rotational eigenvectors and their corresponding eigenvalues for the 3^G subfactor planar algebras. We prove the conjecture in the case when |G| is odd. To do so, we find an action of G on the reduced subfactor planar algebra at f^{(2)}, which is obtained from shading the planar algebra of the even half. We also show that this reduced subfactor planar algebra is a Yang-Baxter planar algebra.

LWE/RLWE-based cryptosystems require sampling error term from discrete Gaussian distribution. However, some existing samplers are somehow slow under certain circumstances therefore efficiency of such schemes is restricted. In this paper, we introduce a more efficient discretized Gaussian sampler based on ziggurat sampling algorithm. We also analyze statistical quality of our sampler to prove that it can be adopted in LWE/RLWE-based cryptosystems. Compared with ziggurat-based sampler by Buchmann et al. related samplers by Peikert, Ducas et al. and Knuth-Yao, our sampler achieves more than 2x speedup when standard deviation is large. This can benefit constructions rely on noise flooding (e.g., homomorphic encryption). We also present two applications: First, we use our sampler to optimize the RLWE-based authenticated key exchange (AKE) protocol by Zhang et al. We achieve 1.14x speedup on total runtime of this protocol over major parameter choices. Second, we give practical post-quantum Transport Layer Security (TLS) ciphersuite. Our ciphersuite inherits advantages from TLS and the optimized AKE protocol. Performance of our proof-of-concept implementation is close to TLS v1.2 ciphersuites and one post-quantum TLS construction.

In this paper, we show that the signal function used in Ring-Learning with Errors (RLWE) key exchange could leak information to find the secret s of a reused public key p = as+2e. This work is motivated by an attack proposed in [1] and gives an insight into how public keys reused for long term in RLWE key exchange protocols can be exploited. This work specifically focuses on the attack on the KE protocol in [2] by initiating multiple sessions with the honest party and analyze the output of the signal function. Experiments have confirmed the success of our attack in recovering the secret.

By analysing the uniform attractor for multi-valued processes, we study the long-time behaviour of the solutions of a model of non-autonomous porous-medium equations. The result is obtained by using the <i>a priori</i> estimates and suitable compactness arguments.