为了提高网络系统中常用的用户名/密码身份鉴别系统的安全性,提出基于用户击键特征的二次身份鉴别方案。该方案在现有的用户名/密码认证方案上加入基于条件随机场的击键特征建模和识别步骤,并在理论上对该方案作了可行性、安全性、效率和准确率分析。通过使用公开数据的实验发现,该方案在提高原方案安全性的同时,具有效率高、准确率高的特点。通过与其他基于相同公开数据的实验结果对比发现,条件随机场模型在击键特征识别领域具有很好的识别效果。
Based on the widely used username /password authentication system,this paper proposed a two-factor authentication system using keystroke dynamics. The scheme embedded keystroke dynamics modeling and identification into the username /password authentication system. Its feasibility,security,efficiency and accuracy were theoretically analyzed. Experiments using publicly available keystroke dynamics datasets show that it has great efficiecy and accuracy with improved security. Performance evaluation with different models using the same public datasets indicates that conditional random field is an effective model for keystroke dynamics.