|
所属 |
工学教育研究部 工学科情報通信工学プログラム担当 |
|
職名 |
特別助教 |
論文 【 表示 ／ 非表示 】
-
Okazaki N., Usuzaki S., Waki T., Kawagoe H., Park M., Yamaba H., Aburada K.
Future Internet 16 ( 8 ) 2024年8月
掲載種別：研究論文（学術雑誌） 出版者・発行元：Future Internet
We propose a detection system incorporating a weighted voting mechanism that reflects the vote’s reliability based on the accuracy of each detector’s examination, which overcomes the problem of cooperative detection. Collaborative malware detection is an effective strategy against zero-day attacks compared to one using only a single detector because the strategy might pick up attacks that a single detector overlooked. However, cooperative detection is still ineffective if most anti-virus engines lack sufficient intelligence to detect zero-day malware. Most collaborative methods rely on majority voting, which prioritizes the quantity of votes rather than the quality of those votes. Therefore, our study investigated the zero-day malware detection accuracy of the collaborative system that optimally rates their weight of votes based on their malware categories of expertise of each anti-virus engine. We implemented the prototype system with the VirusTotal API and evaluated the system using real malware registered in MalwareBazaar. To evaluate the effectiveness of zero-day malware detection, we measured recall using the inspection results on the same day the malware was registered in the MalwareBazaar repository. Through experiments, we confirmed that the proposed system can suppress the false negatives of uniformly weighted voting and improve detection accuracy against new types of malware.
DOI： 10.3390/fi16080259
-
Proposal of a DDoS Attack Detection Method Using the Communication Interval 査読あり
Iwasa K., Usuzaki S., Aburada K., Yamaba H., Katayama T., Park M., Okazaki N.
Lecture Notes in Electrical Engineering 1114 LNEE 165 - 174 2024年
掲載種別：研究論文（学術雑誌） 出版者・発行元：Lecture Notes in Electrical Engineering
As the scale of Distributed Denial of Service (DDoS) attacks has been escalating in recent years, the need for real-time detection of attacks has increased. Existing intrusion detection systems (IDSs) perform detection with a fixed window size (assumed to be in hours). In previous research, attack detection was performed by preparing windows of multiple sizes, selecting the appropriate window based on the state of the data, and using features learned in advance for that window size. Although this method yielded a high DDoS attack detection rate of 98.30%, it exhibited a considerable false-positive rate of 7.37%. The proposed method measures the communication intervals of identical packets within the window, identified as attack-related in the previous survey, and classifies those packets with an average communication interval below a set threshold as attacks. The experiment resulted in a 50.2% decrease in the false-positive rate.
-
Usability improvement in color constancy CAPTCHA 査読あり
Usuzaki, Shotaro, Yihan, Wang, Aburada, Kentaro, Yamaba, Hisaaki, Takatsuka, Kayoko, Katayama, Tetsuro, Park, Mirang, Okazaki, Naonobu
IEICE Communications Express 2024年
掲載種別：研究論文（学術雑誌）
-
Proposal of Fill in the Missing Letters CAPTCHA Using Associations from Images 査読あり
Yamaba H., Mustaza M.N.F.B., Usuzaki S., Aburada K., Mukunoki M., Park M., Okazaki N.
Lecture Notes in Electrical Engineering 1114 LNEE 206 - 217 2024年
掲載種別：研究論文（学術雑誌） 出版者・発行元：Lecture Notes in Electrical Engineering
This paper proposes a new fill in the missing letters type CAPTCHA using associations from images. Many web sites have adopted CAPTCHA to prevent bots and other automated programs from malicious activities such as posting comment spam. Text-based CAPTCHA is the most common and earliest CAPTCHA. But as optical character recognition (OCR) technology has improved, the intensity of distortions that must be applied to a CAPTCHA for it to remain unrecognizable by OCR has increased. This has reached a point where humans are having difficulty recognizing CAPTCHA text. The idea of the proposed CAPTCHA asks users to spell a word by filling some blanks. Since the number of shown letters are few, it is difficult to answer the correct word. But one or more images that can be used as hints to guess what is the answer word are also shown to the users. A series of experiments was carried out to evaluated the performance of the proposed CAPTCHA. First, a computer program was developed with various software languages for the usability evaluation. The system was used for the experiments to find the suitable parameters of the CAPTCHA such as numbers of letters that will be disclosed, position of disclosed letters. Next, security evaluation experiments were carried out using the system under the obtained parameters. The results of the experiments showed that the performance and limitation of the proposed CAPTCHA.
-
Study of an Image-Based CAPTCHA that is Resistant to Attacks Using Image Recognition Systems 査読あり
Nishikawa S., Usuzaki S., Aburada K., Yamaba H., Katayama T., Park M., Okazaki N.
Lecture Notes in Electrical Engineering 1114 LNEE 175 - 184 2024年
掲載種別：研究論文（学術雑誌） 出版者・発行元：Lecture Notes in Electrical Engineering
In today’s digital age, image-based CAPTCHAs are increasingly vulnerable to attacks using annotation services, which tag images and classify images according to their contents, or reverse image search services. To prevent such attacks, an image-based CAPTCHA was proposed that takes advantage of the fact that humans can correctly recognize images containing many discontinuous points, while existing image recognition systems misrecognize them. However, this CAPTCHA proved susceptible to attacks using noise reduction filters. The objective of the present study is to create a CAPTCHA using images that are resistant to such filters. Images used in the new CAPTCHA were realized by increasing the proportion of lines forming discontinuous surfaces in images. Experimental results demonstrated a human recognition rate of 95.8%, with the image recognition systems successfully identifying only one image overall. Moreover, when a noise reduction filter was applied, the recognition rate was lower than those reported in previous studies.
MISC 【 表示 ／ 非表示 】
-
Usuzaki S., Aburada K., Yamaba H., Katayama T., Mukunoki M., Park M., Okazaki N.
Artificial Life and Robotics 27 ( 1 ) 2022年2月
掲載種別：速報，短報，研究ノート等（学術雑誌） 出版者・発行元：Artificial Life and Robotics
In the original publication of the article, on Figs. 4 and 5, “CNN” should read as “MLP”. In addition, under the section “4.2 Human and machine success rate”, on the paragraph “For evaluation of the machine success …”, in the following sentence “In the attack experiment, we applied the famous color …” CNN (Convolutional Neural Network) should be corrected as MLP (Multilayer perceptron). The correct sentence should read as “In the attack experiment, we applied the famous color constancy algorithm, Gray-World, Max-RGB, Gray-Edge [11], Second Derivative Gray-Edge, and MLP (Multilayer perceptron) to images that were saved in the experiment for the human success rate
科研費（文科省・学振・厚労省）獲得実績 【 表示 ／ 非表示 】
-
透過的反射による人間の映り込みを高精度に検出するSNS投稿画像解析手法の開発
研究課題/領域番号：23H05387 2023年04月 - 2024年03月
独立行政法人日本学術振興会 科学研究費補助金 奨励研究
担当区分：研究代表者