Pengesahan menggunakan kata laluan adalah cara yang paling luas digunakan untuk mengenalpasti identiti individu. Namun begitu, cara ini didapati mempunyai banyak kelemahan. Kata laluan memainkan peranan seperti kunci; sesiapa yang mempunyai ia dapat masuk ke dalam sistem. Tambahan pula, kata laluan mudah dipecah masuk, diteka, dicuri, dan dikongsi bersama. Untuk meminimumkan risiko pencerobohan, biometrik penaipan boleh digunakan untuk menambahbaik system kata laluan biasa. Biometrik penaipan mengenalpasti identiti individu berdasarkan cara seseorang menaip di atas papan kekunci. Terdapat penyelidikan yang menggunakan ciri-ciri pemasaan semasa menaip untuk mengenalpasti identiti seseorang. Dalam projek ini, tekanan semasa menaip (tekanan jari di atas papan kekunci) digunakan, dan prestasi dibandingkan dengan teknik penggunaan ciri-ciri pemasaan. Projek ini juga menyelidik penggunaan kombinasi kedua-dua ciri tekanan dan ciri pemasaan. Satu papan kekunci khas yang peka terhadap tekanan telah direkabentuk untuk mengesan tekanan jari semasa menaip. Satu antaramuka pengguna digunakan untuk mengumpul data daripada 100 pengguna. Semua pengguna diminta untuk menggunakan kata laluan yang sama. Tiga cara klasifikasi telah digunakan, iaitu Logistic Regression (LR), Multilayer Perceptron (MLP), dan rangkaian neural Fuzzy ARTMAP (FAM). Keputusan agak menggalakkan, dengan ketepatan setinggi 93.9% didapati dengan menggunakan FAM. Keputusan yang lebih baik didapati dengan menggunakan masa di antara dua penekanan kekunci berturut-turut, berbanding dengan penggunaan tekanan semasa menaip. Tetapi jikalau kedua-dua teknik digabung bersama, keputusan yang lebih baik diperolehi, dengan 0.87% False Acceptance Rate (FAR) dan 4.4% False Rejection Rate (FRR). Keputusan eksperimen-eksperimen yang dijalankan menunjukkan penggunaan tekanan semasa menaip dapat menambah ketepatan kepada system pengesahan biometrik penaipan.
_________________________________________________________________________________________
Password authentication is the most prevalently used identification system in today’s cyber world. In spite of the popularity of this approach there are many inherent flaws. The password plays the role as the key to a lock; anyone who has it can gain successful access. Additionally, passwords can be easily cracked, guessed, stolen or deliberately shared. To minimize the risk of intrusion, keystroke dynamics can be used to complement this popular authentication method. As the name implies, it is an automated biometric method that analyzes the way a person types on a keyboard. There have been a lot of studies on using keystroke timing characteristics to verify the identity of a user. In this project keystroke pressure (the amount of force exerted on each key pressed) was employed, and its performance was compared with that of the conventional keystroke timings-based technique. The project also investigated the use of combined keystroke pressure and latency for the identification process. In order to measure the forces exerted during typing, a pressure-sensitive keyboard system was developed. A user interface that simulates actual login environment was used to collect data from 100 users. All users were requested to enter the same password. Three different classification methods were applied, namely Logistic Regression (LR), Multilayer Perceptron (MLP), and Fuzzy ARTMAP (FAM) neural networks. The results were very encouraging, with a maximum accuracy rate of 93.9% achieved by using FAM. Keystroke latency gave better results than keystroke pressure, but using both techniques together yielded the best results, with False Acceptance Rate (FAR) of 0.87% and False Rejection Rate (FRR) of 4.4%. The experimental results demonstrated that the proposed methods are promising, and that the keystroke pressure is a viable and practical way to add more security to conventional typing biometrics authentication system.
Development of a pressure-based typing biometrics system for user authentication_Loy Chen Change_E3_2005_NI