Searcharxiv⌕ Search

arXiv subjects

Ryohei Suzuki

Publications and source records attributed to Ryohei Suzuki.

3 recordsLinked to original sources
arXiv2023

Applicability of Trust Management Algorithm in C2C services

The emergence of Consumer-to-Consumer (C2C) platforms has allowed consumers to buy and sell goods directly, but it has also created problems, such as commodity fraud and fake reviews. Trust Management Algorithms (TMAs) are expected to be a countermeasure to detect fraudulent users. However, it is unknown whether TMAs are as effective as reported as they are designed for Peer-to-Peer (P2P) communications between devices on a network. Here we examine the applicability of `EigenTrust', a representative TMA, for the use case of C2C services using an agent-based model. First, we defined the transaction process in C2C services, assumed six types of fraudulent transactions, and then analysed the dynamics of EigenTrust in C2C systems through simulations. We found that EigenTrust could correctly estimate low trust scores for two types of simple frauds. Furthermore, we found the oscillation of trust scores for two types of advanced frauds, which previous research did not address. This suggests that by detecting such oscillations, EigenTrust may be able to detect some (but not all) advanced frauds. Our study helps increase the trustworthiness of transactions in C2C services and provides insights into further technological development for consumer services.

cs.CR↗
arXiv2019

Spatially Controllable Image Synthesis with Internal Representation Collaging

We present a novel CNN-based image editing strategy that allows the user to change the semantic information of an image over an arbitrary region by manipulating the feature-space representation of the image in a trained GAN model. We will present two variants of our strategy: (1) spatial conditional batch normalization (sCBN), a type of conditional batch normalization with user-specifiable spatial weight maps, and (2) feature-blending, a method of directly modifying the intermediate features. Our methods can be used to edit both artificial image and real image, and they both can be used together with any GAN with conditional normalization layers. We will demonstrate the power of our method through experiments on various types of GANs trained on different datasets. Code will be available at https://github.com/pfnet-research/neural-collage.

cs.CV↗