Continuous User Authentication Based on Deep Learning Neural Networks

View Statistics Email Alert RSS Feed

Information

Details

Project title

Code/計畫編號

MOST107-2221-E019-018

Translated Name/計畫中文名

運用深度學習之延續性使用者身份認證機制

Project Coordinator/計畫主持人

Pei-Yih Ting

Funding Organization/主管機關

National Science and Technology Council

Department/Unit

Department of Computer Science and Engineering

Website

https://www.grb.gov.tw/search/planDetail?id=12667439

Year

2018

Start date/計畫起

01-08-2018

Expected Completion/計畫迄

31-07-2019

Bugetid/研究經費

399千元

ResearchField/研究領域

資訊科學--軟體

本計畫將運用深度學習的方法來探討各種生理以及行為特徵驗證使用者身份的效能，並且建立一套強健的「延續性使用者身份認證」機制，在不影響使用者操作的前提下以較低的裝置運算能力及較高的準確性不斷地確認使用者的身份，提昇各種應用系統以及移動式裝置的安全性，以高維度的神經元整合多種感測器取得的資訊來降低特徵擷取的困難度，藉由確認多種特徵來提昇機制的可信賴度，擬探討的特徵包括三軸重力加速度感應器及陀螺儀，鍵盤與滑鼠操作習慣，深度相機取得之臉部、耳朵、手指、手掌、手勢之3D資訊，穿戴式感測裝置監測的心跳、血壓、血氧、體溫等資訊，以及環境資訊如Wifi及NFC室內定位資訊。計畫裡也將探討幾種合併這些驗證的方法。在各種行動裝置、智慧型穿戴裝置、以及物聯網裝置的產值與市場規模快速擴大、雲端應用服務以及雲端儲存機制逐漸普遍化之際，「延續性使用者身份認證」機制的重要性日漸提昇，本項研究中各個子項目都可以對相關的產品產生極大的加值作用，催化產業界對於產品不同功能的研發，對於行動裝置、穿戴裝置、物聯網產業的轉型與技術深化有很大的影響，對於各種網路金融與電子商業服務的普及化極為關鍵。 Many physiological and behavioral biometric features, including face, fingerprint, palm print, ear, typing and mouse dynamics, were studied extensively to verify the identity of a user through machine learning techniques. Nowadays, mobile and wearable devices are equipped with more powerful sensors that can monitor many more physiological traces of a user. These traces, although their features might not be easy to extract for classification or verification, can be processed directly by deep neural networks for the development of a continuous user authentication mechanism to enhance the security of a network application or a mobile device. In this proposal, we use deep learning network to process high dimensional input vectors directly, abstract them in its own way, and measure the likelihood according to its pre-trained model. We plan to investigate measurements like gyroscope, accelerometer, keyboard/mouse dynamics/behaviors, 3D face, ear, finger, palm, and gestures from a TOF depth camera, heart rate, blood pressure, blood oxygen, body temperature, and environmental features like signal intensity of Wifi APs. We also like to investigate the fusion of decisions from multiple neural networks. The development of continuous user authentication techniques will be very crucial in successful promotion of mobile device usages and network/cloud e-commerce applications.

Keyword(s)

延續性身份認證
深度學習
機器學習
生理以及行為特徵
雲端服務
行動裝置
continuous user authentication
deep learning
machine learning
physiological and behavioral features
cloud services
mobile devices

DSpace CRIS

Continuous User Authentication Based on Deep Learning Neural Networks

Details

Description