Ryuya Uda

Bio: Ryuya Uda is an academic researcher from Tokyo University of Technology. The author has contributed to research in topics: CAPTCHA & Password. The author has an hindex of 6, co-authored 62 publications receiving 184 citations.

A method of appliance detection based on features of power waveform

Abstract: We propose a method to detect the model, location and activity of a conventional home electric appliance. Waveforms of current consumed by appliances vary according to their configurations and activity. We define feature parameters for detecting the status of appliances. A current detector, microcomputer and transmitter are equipped in a power outlet in order to measure consumed current, calculate the feature parameters, and transmit the results to a home server. Feature parameters of appliances in the home are learned and stored in a home server in advance. The home server compares the feature parameters of known appliances with the received feature parameters to detect an appliance's model and activity. User could control appliances from out of the house via the Internet.

Static analysis with paragraph vector for malware detection

Abstract: Malware damages computers and the threat is a serious problem. Malware can be detected by pattern matching method or dynamic heuristic method. However, it is difficult to detect all new malware subspecies perfectly by existing methods. In this paper, we propose a new method which automatically detects new malware subspecies by static analysis of execution files and machine learning. The method can distinguish malware from benignware and it can also classify malware subspecies into malware families. We combine static analysis of execution files with machine learning classifier and natural language processing by machine learning. Information of DLL Import, assembly code and hexdump are acquired by static analysis of execution files of malware and benignware to create feature vectors. Paragraph vectors of information by static analysis of execution files are created by machine learning of PV-DBOW model for natural language processing. Support vector machine and classifier of k-nearest neighbor algorithm are used in our method, and the classifier learns paragraph vectors of information by static analysis. Unknown execution files are classified into malware or benignware by pre-learned SVM. Moreover, malware subspecies are also classified into malware families by pre-learned k-nearest. We evaluate the accuracy of the classification by experiments. We think that new malware subspecies can be effectively detected by our method without existing methods for malware analysis such as generic method and dynamic heuristic method.

Classification of XSS Attacks by Machine Learning with Frequency of Appearance and Co-occurrence

Abstract: Cross site scripting (XSS) attack is one of the attacks on the web. It brings session hijack with HTTP cookies, information collection with fake HTML input form and phishing with dummy sites. As a countermeasure of XSS attack, machine learning has attracted a lot of attention. There are existing researches in which SVM, Random Forest and SCW are used for the detection of the attack. However, in the researches, there are problems that the size of data set is too small or unbalanced, and that preprocessing method for vectorization of strings causes misclassification. The highest accuracy of the classification was 98% in existing researches. Therefore, in this paper, we improved the preprocessing method for vectorization by using word2vec to find the frequency of appearance and co-occurrence of the words in XSS attack scripts. Moreover, we also used a large data set to decrease the deviation of the data. Furthermore, we evaluated the classification results with two procedures. One is an inappropriate procedure which some researchers tend to select by mistake. The other is an appropriate procedure which can be applied to an attack detection filter in the real environment.

Text-Based CAPTCHA Using Phonemic Restoration Effect and Similar Sounds

Abstract: In Recent years, bot (robot) program has been one of the problems on the web. Some kinds of the bots acquire accounts of web services in order to use the accounts for SPAM mails, phishing, etc. CAPTCHA (Completely Automated Public Turing Test to Tell Computers and Humans Apart) is one of the countermeasures for preventing bots from acquiring the accounts. Text-based CAPTCHA is especially implemented on almost all famous web services. However, CAPTCHA faces a problem that evolution of algorithms for analysis of printed characters disarms text-based CAPTCHA. Of course, stronger distortion of characters is the easiest solution of the problem. However, it makes recognition of characters difficult not only for bots but also for human beings. Therefore, in this paper, we propose a new CAPTCHA with higher safety and convenience. Especially, we focus on the human abilities of phonemic restoration and recognition of similar sounds, and adopt the abilities in the propose CAPTCHA. The proposed CAPTCHA makes machinery presumption difficult for bots, while providing easy recognition for human beings.

ADEC: Assured Deletion and Verifiable Version Control for Cloud Storage

Abstract: Although cloud storage offers a number of attractive features, it also raises various security concerns. In particular, cloud storage of sensitive information, such as medical or trade records, requires measures to be taken not only for privacy protection, but also for assured deletion and verifiable version control. Even though there are methods for realizing both assured deletion and version control in cloud storage, they cannot guarantee the integrity of files and their versioning order. Therefore, we propose a system referred to as ``Assured Deletion and verifiable version Control (ADEC)'', which is implemented as a virtual file system capable of taking snapshots in cloud storage. The main ideas behind ADEC are a scheme for generating chained keys and a scheme for verifiable version control with hysteresis signatures. This method can be used not only for protection of privacy, but also for assured deletion of specific versions and all older versions in a single operation. In addition, the integrity of each file version can be verified with the hysteresis signature scheme, which makes it impossible to implement rollback and reordering attacks.

Unsupervised disaggregation of low frequency power measurements

Abstract: Fear of increasing prices and concern about climate change are motivating residential power conservation efforts. We investigate the effectiveness of several unsupervised disaggregation methods on low frequency power measurements collected in real homes. Specifically, we consider variants of the factorial hidden Markov model. Our results indicate that a conditional factorial hidden semi-Markov model, which integrates additional features related to when and how appliances are used in the home and more accurately represents the power use of individual appliances, outperforms the other unsupervised disaggregation methods. Our results show that unsupervised techniques can provide perappliance power usage information in a non-invasive manner, which is ideal for enabling power conservation efforts.

Design and implementation of a high-fidelity AC metering network

Abstract: We present the architecture, design, and preliminary evaluation of ACme, a wireless sensor and actuator network for monitoring AC energy usage and controlling AC devices in a large and diverse building environment The ACme system consists of three tiers: the ACme node which provides a metering and control interface to a single outlet, a network fabric which allows this interface to be exported to arbitrary IP endpoints, and application software that uses this networked interface to provide various power-centric applications The ACme node integrates an Epic core module with a dedicated energy metering IC to provide real, reactive, and apparent power measurements, with optional control of an attached load The network comprises a complete IPv6/6LoWPAN stack on every node and an edge router that connects to other IP networks The application tier receives and stores readings in a database and uses a web server for visualization Nodes automatically join the IPv6 subnet after being plugged in, and begin interactions with the application layer We evaluate our system in a preliminary green building deployment with 49 nodes spread over several floors of a Computer Science Building and present energy consumption data from this preliminary deployment

A platform for ubiquitous sensor deployment in occupational and domestic environments

Abstract: In this paper, we introduce the "Plug" sensor network, a ubiquitous networked sensing platform ideally suited to broad deployment in environments where people work and live. The backbone of the Plug sensor network is a set of 35 sensor-, radio-, and computation-enabled power strips distributed throughout the third floor of the MIT Media Lab. A single Plug device fulfills all the functional requirements of a normal power strip (i.e., four 120 V, 60 Hz electrical outlets; surge protector circuit; standard electrical connector to a US-style wall socket), and can be used without special training. Additionally, each Plug has a wide range of sensing modalities (e.g., sound, light, electrical current and voltage, vibration, motion, and temperature) for gathering data about how it is being used and its nearby environment. To our knowledge, the Plug sensor network is the first to embody the idea of designing sensor nodes to seamlessly become a part of their environment, rather than play the role of alien, if unobtrusive, observers. We argue this design principle is essential for sensor networks to succeed in the realm of ubiquitous computing. In this paper, we present an overview of the Plug hardware and software architectures, look at specific usage scenarios of a single Plug, and show example data taken across the entire Plug network to give a sense of the pulse of the building over a span of days. Finally, we present ongoing work interfacing heterogeneous devices with the Plug network for a variety of applications and discuss possible future work.

A Survey on Intrusive Load Monitoring for Appliance Recognition

Abstract: Electricity load monitoring of appliances has become an important task considering the recent economic and ecological trends. In this game, machine learning has an important part to play, allowing for energy consumption understanding, critical equipment monitoring and even human activity recognition. This paper provides a survey of current researches on Intrusive Load Monitoring (ILM) techniques. ILM relies on low-end electricity meter devices spread inside the habitations, as opposed to Non-Intrusive Load Monitoring (NILM) that relies on an unique point of measurement, the smart meter. Potential applications and principles of ILMs are presented and compared to NILM. A focus is also given on feature extraction and machine learning algorithms typically used for ILM applications.

Detecting actuation of electrical devices using electrical noise over a power line

Abstract: Activity sensing in the home has a variety of important applications, including healthcare, entertainment, home automation, energy monitoring and post-occupancy research studies. Many existing systems for detecting occupant activity require large numbers of sensors, invasive vision systems, or extensive installation procedures. Disclosed is an approach that uses a single plug-in sensor to detect a variety of electrical events throughout the home. This sensor detects the electrical noise on residential power lines created by the abrupt switching of electrical devices and the noise created by certain devices while in operation. Machine learning techniques are used to recognize electrically noisy events such as turning on or off a particular light switch, a television set, or an electric stove. The system has been tested to evaluate system performance over time and in different types of houses. Results indicate that various electrical events can be learned and classified with accuracies ranging from 85-90%.