User Authentication Using Wifi Enabled Finger Gesture Detection
Oral Defence Date:
Thursday, July 11, 2019 - 12:00
Assist. Prof. Hao Yue & Prof. Ilmi Yoon
User authentication is one of the essential processes in applications related to Internet of things (IoT) and smart home automation. There has been a reported lack of confidence in IoT commercial products because of all the hacking and attacks on user accounts. Traditional authentication approaches vary from password-based mechanisms to biometric-based mechanisms, which are either insecure due to weak passwords or require users to install/wear extra devices that is inconvenient. To overcome those limitations, this project proposes a device-free authentication mechanism that exploits the prevalent wireless signals in the indoor environment. We collect the data called Channel State Information (CSI) from Wi-Fi enabled devices when the authorized user performs a finger gesture and extract the unique pattern in the CSI data as the fingerprint for the users' finger gesture. We then build a machine learning classifier and train it with the CSI data to recognize the finger gesture from the authorized user and deny unauthorized users. We perform extensive experimentations and the system can achieve an accuracy of 89% in authenticating users.
Computer networks, Network security, Machine learning, Autoencoder, Authentication, IoT, Matlab