論文

In this chapter, we present a device-driver-based computer that realizes the reduction of mode (domain or vertical) switching overheads between user and kernel mode with innovative attributes, including shared keyboards and mice, access-controlled files, and timed files. Experimented results show that old personal computers can revive again with the proposed Driverware technology. The proposed Driverware can improve the CPU resource utilization by three times.

With the rapid advance in digital network, digital libraries, and particularly WWW (World Wide Web) services, we can retrieve many kinds of images on personal and mobile computer anytime and anywhere. At the same time, secure image archiving is becoming a major research area because the serious concern is raised about copyright protection and authority identification in digital media. A more sophisticated technique is required for future multimedia copyright protection. In this chapter we propose a secure image archiving using novel digital-watermarking techniques. Firstly, a nonlinear adaptive system (neural network) is applied for frequency-based digital watermarking. Secondly, we discuss application-oriented watermarking method for GIS image archiving. This chapter is divided into two parts. First section is about the way to apply nonlinear adaptive system for frequency-based image watermarking. We propose a new asymmetric technique employing nonlinear adaptive system trained on frequency domain. Our system uses two public keys to prevent removal attack and archive more fragile watermarking. In embedding, location information of frequency domain, where adaptive system is trained, is binalized, expressed in hexadecimal number, and encrypted in asymmetric cryptosystem. Encrypted location information is embedded in several parts of digital host contents. In generating key, supervised neural networks learn to assign the array of coefficients to teacher signal corresponding to the message to insert. This is one kind of transform-based method to generate public key from private key. In extracting, we use key matrix created by one-way signal processing of adaptive system. Proposal method is tested in still image, and we have empirically obtained the results that the proposal model is functional in implementing more secure and fragile watermarking compared with previous techniques, such as correlation and transform-based asymmetric watermarking. Several experiments are reported to validate the effectiveness of our watermarking method. Second section is about the application of GIS image archiving using digital watermarking technique. Recently, the utilization of GIS (geographical information system) is becoming rapidly pervasive. Consequently, new methodology of archiving and managing images is a pressing problem for GIS users. It is also expected that as the utilization of GIS becomes widely spread, protecting copyright and confidential images will be more important. In this chapter, we propose a three-layer image data format that makes it possible to synthesize two kinds of related images and analysis information in one image data size. To achieve the confidentiality of one hidden image, we apply the private watermarking scheme, where the algorithm is closed to the public. In the proposal model, encoder netlist embedded in the third layer is generated by FOL prover to achieve more secure and less information to decode it, compared with one operation of another block cipher such as RSA. Proposal system users can process two images without the cost of maintaining key and decoding operation.

The rapid increasing of security incidents imposes a great burden on Internet users and system administrators. In this paper we discuss a parallel analysis for lightweight network incident detection using nonlinear adaptive systems. We run AID (anomaly intrusion detection) and MID (misuse intrusion detection) systems in parallel. Two detectors generate binary output misuse = {YES/NO} and anomaly = {YES/NO}. Then, we can determine whether we need to perform network or security operation. We apply clustering algorithm for AID and classification algorithm for MID. The nonlinear adaptive system is trained for running MID and AID in parallel. Proposed parallel system is more lightweight and simple to operate even if the number of incident patterns is increased. Experimental results in the case where false positive is frequently caused show that our method is functional with a recognition rate of attacks less than 10%, while finding the anomaly status. Also, performance evaluation show that proposed system can work with reasonable CPU utilization compared with conventional serial search based system.

Code division multiple access (CDMA) is widely used because of its effectiveness to send multiple signal and confidentiality of career signal. We present a formulation of state-space problem of which solution is directed by redundant reasoning control method for semiheuristic and lightweight DS-CDMA code acquisition. The reasoning of the state-space problem provides us with the way to find a K bit synchronized sequence among K dephased sequences with less calculation cost, compared with serial search and matched filter. In this process, redundancy-restriction method, called weighting strategy, enhances the searching ability of FOL (first order logic) reasoning for the faster and lightweight code acquisition. The combination of weighting strategy and correlator enables us to achieve the peak-detection within K/3 times of calculating inner products and its measurement. Our system is evaluated by the reduced cost of proving state-space problem using weighting strategy and its robustness of using the proposal code acquisition framework. Experiment shows that the proposal method is robust if K/N sequences are grouped with N ranging from 3 to 5.

In this chapter, we propose a new information hiding and extracting method without embedding any information into the target content by using a nonlinear feature extraction scheme trained on frequency domain. The proposed method can detect hidden bit patterns from the content by processing the coefficients of the selected feature subblocks to the trained neural network. The coefficients are taken from the frequency domain of the decomposed target content by frequency transform. The bit patterns are retrieved from the network only with the proper extraction keys provided. The extraction keys, in the proposed method, are the coordinates of the selected feature subblocks and the neural network weights generated by the supervised learning of the neural network. The supervised learning uses the coefficients of the selected feature subblocks as the set of input values, and the hidden bit patterns are used as the teacher signal values of the neural network, which is the watermark signal in the proposed method. With our proposed method, we are able to introduce a watermark scheme with no damage to the target content.

File-system integrity tools (FIT) are commonly deployed host-based intrusion detections (HIDS) tool to detect unauthorized file-system changes. While FIT are widely used, this kind of HIDS has many drawbacks: the intrusion detection is not done in real-time manner, which might render the whole scheme useless if the attacker can somehow take over the system with privileged access in the time between. The administrator also has a lot of problems to keep the base-line database updating. Besides, the database and the FIT itself are vulnerable if the attacker gains local privileged access.This paper presents a novel approach to address the outstanding problems of the current FIT. We propose a design and implementation of a tool named XenFIT for Xen virtual machines. XenFIT can monitor and fires alarms on intrusion in real-time manner, and our approach does not require to create and update the database like in the legacy methods. XenFIT works by dynamically patching memory of the protected machine, so it is not necessary to install any kernel code or user-space application into the protected machines. As a result, XenFIT is almost effortless to deploy and maintain. In addition, thanks to the advantage introduced by Xen, the security polices as well as the detection process are put in a secure machine, so XenFIT is tamper-resistant with attack, even in case the attacker takes over the whole VM he is penetrating in. Finally, if deploying strictly, XenFIT is able to function very stealthily to avoid the suspect of the intruder.

File-system integrity tools (FIT) are commonly deployed to assist forensic investigation after security incidents and as host-based intrusion detections (HIDS) tool to detect unauthorized file-system changes. Basically all the current solutions employ the same tactic: the administrator specifies a list of critical files and directories that needs to be monitored, then uses the FIT to create a base-line database that tracks general parameters about these files. The FIT is then re-run periodically, and if it detects the modifies of the file-system against the information stored in the database, the report on the changed file is generated. However, this strategy is far from perfect: the intrusion detection cannot be done in real-time, which might render the whole scheme useless if the attacker can somehow take over the system with privileged access in the time between. The administrator also has a lot of problems to keep the database updating. Besides, he must do everything he can to protect the database and the FIT itself from compromising by the attacker, which is not an easy task especially if the attacker gains local access. This paper presents a novel approach to address the outstanding problems of the current FIT. We propose a design and implementation of a tool named XenRIM for Xen virtual machines. XenRIM can monitor and fires alarms on intrusion in real-time manner, and our approach does not require to create and update the database like in the legacy methods. As a result, XenRIM is almost effortless to deploy and maintain. Thanks to the advantage introduced by Xen, the detection polices are centralized in a secure virtual machine and resistant to tampering. Even better, if deploying strictly this tool is able to function very stealthily to avoid the suspect of the attacker. Our experimental result demonstrates that XenRIM incurs very low performance overhead (less than 4%), which makes the solution attractive and practical for production systems.

Honeypot is a decoy system to trap attackers, and data capture tool is one of the components of the honeypot architecture. Being used to collect the intruder's activities inside the honeypot, this key component must be able to function as stealthily as possible, so the intruder does not know that he is under watch. Unfortunately Sebek, a de-facto tool for this purpose in the modern honeypot technology, is rather easy to detect, even with unprivileged right access. This paper proposes to use Xen Virtual Machine to deploy honeypot, and takes the advantage introduced by Xen to fix some of the outstanding problems of Sebek. We present a design and implementation of a Xen-based system named Xebek as a solution. While Xebek provides similar features as Sebek does, our system is more "invisible" and harder to defeat. The experimental results also demonstrate that Xebek is more flexible, while the reliability and efficiency are significantly improved over its counterpart.

When we encounter an interesting entity (e.g., a person's name or a geographic location) while reading text, we typically search and retrieve relevant information about it. Entity linking (EL) is the task of linking entities in a text to the corresponding entries in a knowledge base, such as Wikipedia. Recently, EL has received considerable attention. EL can be used to enhance a user's text reading experience by streamlining the process of retrieving information on entities. Several EL methods have been proposed, though they tend to extract all of the entities in a document including unnecessary ones for users. Excessive linking of entities can be distracting and degrade the user experience. In this paper, we propose a new method for evaluating the helpfulness of linking entities to users. We address this task using supervised machine-learning with a broad set of features. Experimental results show that our method significantly outperforms baseline methods by approximately 5.7%-12% F1. In addition, we propose an application, Linkify, which enables developers to integrate EL easily into their web sites.

Recent years, robots are useful at home such as cleaning task or communication tools. So there are a lot of studies about cooperative behavior with robots. In order to realize cooperative tasks with robots, it is necessary that robots estimate human intention from human behavior and act in the context based on the intention. In this research, we construct an agent model that enables coordinated behavior by estimating human intention. We focus on collision avoidance as an example of a simple cooperative behavior. We implemented that a agent has Meta-Strategy model in a virtual environment. We have a collision avoidance experiment between virtual agent and human subject and analyzed subject's behavior. It was confirmed that the agent's behavior can influence the human avoidance behavior from experimental results. By indicating the agent's intention, we consider it is possible to achieve cooperative collision avoidance.

HTTP is absolute essential part of the Internet. However, in the large delayed network such as mobile data network, the performance of HTTP is influenced by 'TCP acknowledgement delay'. In this paper, HTTP over WebSocket proxy system is proposed to defeat the delay. In proposal, HTTP is used to achieve end-to-end communication. Each HTTP request and response is encapsulated into WebSocket packet at the client and server. In this paper, proposed system is implemented on Android smartphone and evaluated delay difference characteristics.

Due to the influence of this aging society, the importance of rehabilitation is increasing in recent years, and new type of rehabilitation, self-exercise method, has been proposed. In order to verify such rehabilitation effect, we measure and compare several rehabilitation effects with fNIRS. We establish an analysis method of fNIRS rehabilitation data, and compare those rehabilitations, and consider what kind of rehabilitation is effective way. In this study, the effectiveness of self-exercise method is suggested.

The rehabilitation technique of the motivative exercise based on Takizawa-Program is suggested to undergo rehabilitation of the increasing number of the hemiplegic patient recently. It is thought that this rehabilitation can reduce the burden of physical therapist and enable that the patients walk on themselves. But it's not become popular because the method of quantitative evaluation of rehabilitation effect is not built. Our research purpose is to examine how to change of brain functional active area by continued Takizawa-Program by hemiplegic patient. Then, we correlated during fNIRS and fMRI by projecting longitudinal plane of fMRI into two dimensional planes. We would like to examine the change of active area around brain functional lesions by projection algorithm.

SystemVerilog assertion (SVA) is a way to express properties that are expected to be true in a design described in Verilog HDL IEEE1364 standard. We have already reported that legitimate peripheral participation (LPP) works very well for the fine grain microprocessor design education on FPGA where the heart of the system is chosen as the way-in which is the first step for the observation in LPP. We have demonstrated its effectiveness on superscalar design education, while the prior pipeline design education failed. The failure was caused by the top down design methodology guided in the education for the pipelining which appeared to be too difficult. Appropriate scheme to observe the heart of the pipelining is needed. We have found that SVA plays a key role where two senior students succeeded to design pipelined RISC having 3 stages and pipelined CISC having 4 stages in 2 months. White box test by using SVA enables the two senior students to observe the heart of the pipelining very effectively.

In this paper, we propose Micro Secure Socket Layer (MSSL) for 8-bit flash micro controller that is about 1.3 Kbytes in code size. We have analyzed and compared various cryptographic protocols in TCP/IP stack for 5$ Micro Server to propose a simple secure layer based on simple handshake processing and encryption. Additionally, we analyzed MSSL performance, compared with the current security protocols and proved that MSSL can secure the connection between Micro Server and Internet. Security implementation for Micro Server, which has very limited size of memory with a small processer, is very difficult and challenging task. Conventional researches have focused mainly on creating small sensor system and Micro Server that has multi-function with low cost and small memory size, but without much security considerations. Crackers can easily access to the sensor nodes or Micro Server. Therefore, minimum security is needed in the ubiquitous sensor networks. The proposed MSSL is very small in code size so it can be implemented and is suitable for small sensors and Micro Server systems.

In this paper, we propose a simple IPSec protocol stack for Micro Server. We proposed an implementation of IPSec protocol stack which is constructed by Encapsulating Security Payload (ESP) protocol with Advanced Encryption Security (AES) encryption scheme, whereas authentication using MD5 algorithm is optional. Researchers have focused on creating a small system composed of sensors and a Micro Server where it has a small sized memory, multi-function, low cost, but without security consideration. The security problem in the Micro Server is a challenging task because of the very limited flash memory. Here, we have implemented the AES function as 2.704 Kbytes and the ESP protocol with this encryption function as 3.822Kbytes of code. Therefore, the proposed method has less than 4Kbytes in code size. Even including the authentication using MD5, the file size is less than 7Kbytes although this is still optional. In our proposed method, we have focused on implementing the encapsulation of the payload and ignored the key exchange procedure to simplify the secure communication compared to conventional IPSec protocol stack.

In this paper, we present a new key generation model for image hashing using neural network, which does not embed any data into the content but is able to extract meaningful data from target image. This model trains artificial neural network to assign predefined code and uses this trained artificial neural network weight and the coordinates of the selected feature sub blocks of target image as keys to extract the predefined code. In this model, the observed output signal from the trained neural network is used as image hash value which distinguishes the target image from other images. The proposed method contributes to secure image hashing for content identification without damaging or losing any detailed data of visual images. The proposed method realizes an application for image authentication, image similarity comparison, verification of image integrity and copyright protection of multimedia contents.

This paper considers a perceptually motivated estimator for single-channel speech enhancement based on statistics and Bayesian estimation in the frequency domain. In detail, we propose a new speech log-spectral amplitude estimator where the cost function is the weighted Euclidean distortion measure of the speech log-spectral amplitude. That cost function is motivated by auditory masking effects of the human hearing system. The statistical assumptions used to develop the proposed estimator are the complex Gaussian distribution and independence of speech, noise discrete Fourier transform coefficients. We evaluate this estimator with speech signals contaminated by various noise sources at different input signal-to-noise ratios and find that it achieves better performance than the well-known minimum mean square error log-spectral amplitude estimator in terms of both noise reduction and speech quality.