Self-Supervised Learning (SSL) has gained traction for its ability to learn rich representations with low labeling costs, applicable across diverse downstream tasks. However, assessing the downstream-task performance remains challenging due to the cost of extra training and evaluation. Existing methods for task-agnostic evaluation also require extra training or hyper-parameter tuning. We propose a novel evaluation metric using large language models (LLMs). By inputting discrete token sequences and minimal domain cues derived from SSL models into LLMs, we obtain the mean log-likelihood; these cues guide in-context learning, rendering the score more reliable without extra training or hyperparameter tuning. Experimental results show a correlation between LLM-based scores and automatic speech recognition task. Additionally, our findings reveal that LLMs not only functions as an SSL evaluation tools but also provides inference-time embeddings that are useful for speaker verification task.

MultiPath TCP (MP-TCP) allows the use of multiple paths between two end hosts for a single data transmission, extending the capabilities of SinglePath Transmission Control Protocol (SP-TCP). AIMD congestion control algorithm can be implemented on each subpath of the MP-TCP sender. However, the characteristics of the AIMD-type window flow control depend on the control parameters (α, β). There have been some guidelines proposed to select proper control parameters (α, β) that can achieve fair bandwidth sharing between SP-TCP and MP-TCP senders using existing MP-TCP congestion algorithms. However, current guidelines do not offer a solution that can simultaneously increase the throughput of an MP-TCP sender and ensure fairness between MP-TCP and SP-TCP senders. To address this issue, this study proposes a control parameter setting for an MP-TCP sender that can maximize the sender’s throughput and ensure fairness between SP-TCP and MP-TCP senders. We derive a fluid model of a network that includes SP-TCP and MP-TCP senders, describing the relationship between control parameters, the aggregate throughput of an MP-TCP sender, and the packet loss rate of a router.

Information-Centric Networking (ICN) has gained attention as one of the next-generation internet architectures that focuses on the data being transmitted rather than the hosts transmitting it. Due to the differences between ICN and TCP/IP networks, it is not possible to evaluate the performance of ICN using network simulators designed for TCP/IP. A number of studies have been conducted to develop ICN network simulators. However, further acceleration of ICN network simulators is expected to enable large-scale ICN network performance evaluation. In this paper, we analyze the performance bottleneck of the flow-level ICN simulator called FICNSIM (Fluid-based ICNSIMulator) by profiling its performance using the Julia language source code. Specifically, we identify the processing that is causing the performance bottleneck of FICNSIM and investigate the scalability of FICNSIM with respect to network scale.

We propose an algorithm to automatically restrict the step size in the iterative optimization process with an application to adversarial attacks on speaker verification models. The proposed algorithm dynamically determines a subspace with a restriction radius r to which the Taylor approximation is applied at each iteration and then solves a linear problem within the subspace by using the projected gradient method. In experiments, we demonstrate adversarial attacks on three speaker verification models: i-vectors, SE-ResNet-34, and ECAPATDNN. We show that the degree of adversarial perturbations generated by the proposed algorithm is smaller than that generated by the conventional attack method.

In past research, we proposed an authentication method that combines actions with spatiotemporal information such as location, time, and distance. With the method, a user succeeds in authentication when he/she performs preset actions such as pushing button n times on preset intervals defined by spatiotemporal information. In this paper, we improve the authentication method using a partial matching method. We propose two kinds of partial matching methods for pushing button and interval. A type I method assumes the number of pushing button is sometimes less than preset count, but the number never exceeds it, and a user never pushes the button out of preset areas. A type II method assumes the number of pushing button is less or more than preset count occasionally, and a user pushes the button out of preset areas. We showed how to calculate FAR when the type I or II is applied. In the experiment, we compared the type I and II methods with a conventional method to evaluate their security. As a result, the type I method improved false acceptance rate (FAR) from 0.097% to 0.053%. The type II method improved FAR from 0.097% to 0.035%.

In many software development projects, people tend to repeat same mistakes due to lack of shared knowledge from past experiences. Generally, it is very difficult to manually find out valuable phenomena from huge data. Invisible context, which cannot be known directly from software documents or formal reports, is an important factor to these difficulties. We propose a new method to find contexts based on analysis to email archives in a project repository. In this method, we first apply natural language processing to extract keywords from email messages. Next, similarities among the messages are calculated based on the extracted keywords, and the messages are classified into clusters according to the similarities. The clustering result can be presented with other information such as code growth graph or schedule charts. This method is implemented as an extension to the Project Replayer, a tool to review past project data. Pilot analysis confirms that a researcher could grasp important contexts of failures in actual projects using the Project Replayer.