Various services ranging from finance to public services are digitalized in recent years for higher efficiency and user convenience. With this service digitalization, the need for identifying and authenticating users is increasing. Amongst the user authentication methods, biometric authentication is spreading as it does not require the user to remember a password or to have a specific token. As a more convenient authentication method, research is also being conducted on unconscious authentication using smartphones’ movement history. In this paper, we propose location history-based implicit user authentication acquired through GPS-equipped mobile devices. This method enables hands-free user authentication just by having a mobile device. However, location data are sensitive information that needs to be secured from the risk of location data leakage. By using the template protection technique, location data can be transformed so that the original location data cannot be recovered while enabling authentication. However, it has a trade-off between security and accuracy and remains as a problem to be solved. This paper proposes a new location history matching method based on Modified Weighted Jaccard Coefficient. Then it extends it to template protected location history authentication by presenting a new template protection technique using b-Bit MinHash. Our experimental results show that our proposed location matching method achieves practical accuracy compared with the conventional location history matching method. Furthermore, our template-protected location authentication has comparable accuracy to unprotected matching.

With the growth of image on the web, research on hashing which enables high-speed image retrieval has been actively studied. In recent years, various hashing methods based on deep neural networks have been proposed and achieved higher precision than the other hashing methods. In these methods, multiple losses for hash codes and the parameters of neural networks are defined. They generate hash codes that minimize the weighted sum of the losses. Therefore, an expert has to tune the weights for the losses heuristically, and the probabilistic optimality of the loss function cannot be explained. In order to generate explainable hash codes without weight tuning, we theoretically derive a single loss function with no hyperparameters for the hash code from the probability distribution of the images. By generating hash codes that minimize this loss function, highly accurate image retrieval with probabilistic optimality is performed. We evaluate the performance of hashing using MNIST, CIFAR-10, SVHN and show that the proposed method outperforms the state-of-the-art hashing methods.

A cancelable biometric scheme called correlation-invariant random filtering (CIRF) is known as a promising template protection scheme. This scheme transforms a biometric feature represented as an image via the 2D number theoretic transform (NTT) and random filtering. CIRF has perfect secrecy in that the transformed feature leaks no information about the original feature. However, CIRF cannot be applied to large-scale biometric identification, since the 2D inverse NTT in the matching phase requires high computational time. Furthermore, existing biometric indexing schemes cannot be used in conjunction with template protection schemes to speed up biometric identification, since a biometric index leaks some information about the original feature. In this paper, we propose a novel indexing scheme called “cancelable indexing” to speed up CIRF without losing its security properties. The proposed scheme is based on fast computation of CIRF via low-rank approximation of biometric images and via a minimum spanning tree representation of low-rank matrices in the Fourier domain. We prove that the transformed index leaks no information about the original index and the original biometric feature (i.e., perfect secrecy), and thoroughly discuss the security of the proposed scheme. We also demonstrate that it significantly reduces the one-to-many matching time using a finger-vein dataset that includes six fingers from 505 subjects.

An information management system using analyzing document structure is presented. The purpose is simultaneous management of information in various paper and electronic documents. The system contains image document analysis, PDF document analysis, and HTML document analysis. The two applications are presented and the developed prototypes are described. One application is document summarization. The other application is table understanding to correlate data to items.

Constellation matching is a simple but useful image processing method. The method makes it possible to recognise the image or the document by abstracting its essential targets like fixed stars in a constellation. The method is composed of two parts: (i) template matching to extract targets such as essential partial parts in the image or important words in the document; (ii) point pattern matching to examine the positional or the semantic relationship between those targets. This paper firstly describes the mathematical analysis which gives us the relationship between the number of targets and the recognition rate, and that is helpful to estimate the limit of the system ability and to design system parameters. Secondly, the application to document form identification using keywords detected by character recognition is discussed. An evaluation using 139 different document forms gave us the result of 98% correct identification and 2% rejection rate.