Keyboard and mouse based behavioral biometrics to enhance password-based login authentication using machine learning model
In one approach, a method includes: receiving a login event input from a user, the login event input being associated with a session of the user logging into an account; accessing a machine learning model; and authenticating, with the machine learning model, the user for the account, based at least in part on the login event input. In examples, the login event input comprises one or more items of biometric data associated with the user, an item of the one or more items of biometric data associated being generated by interaction of the user with an input device for logging into the account, and the interaction communicating a login credential of the user. In examples, an item of the one or more items of biometric data associated with the user is keyboard event-related biometric data, or mouse event-related biometric data.
1. A method, comprising:
receiving a login event input from a user, the login event input being associated with a session of the user logging into an account;
accessing a machine learning model;
authenticating, with the machine learning model, the user for the account, based at least in part on the login event input;
the login event input comprises one or more items of biometric data associated with the user;
an item of the one or more items of biometric data associated with the user is generated by interaction of the user with an input device for logging into the account, the interaction communicating a login credential of the user;
an item of the one or more items of biometric data associated with the user is keyboard event related biometric data, or mouse event related biometric data;
the machine learning model is trained by a plurality of login event inputs associated with sessions of the user logging into the account;
the machine learning model has been trained by a predetermined threshold number of sessions of the user logging into the account; and
a vector of mouse features is fused with a vector of keyboard features to generate a vector of login features.
2. The method of claim 1 , wherein an item of the one or more items of biometric data is obtained by monitoring the user interaction with an input device during the session.
3. The method of claim 1 , wherein the keyboard event related biometric data comprises a set of features associated with keystrokes of the user during the session.
4. The method of claim 1 , wherein the mouse event related biometric data comprises a set of features associated with dynamics of a mouse movement of the user during the session.
5. The method of claim 1 , wherein features of biometric data is generated independently of an environment of the user during the sessions.
6. The method of claim 1 , wherein the machine learning model is trained by a different vector based on a plurality of historical user login sessions.
7. The method of claim 1 , wherein the machine learning model is based on a random forest classifier.
8. The method of claim 1 , wherein the machine learning model is trained in a biased manner that is tolerative of false positive authentication results and sensitive to false negative authentication results.
9. The method of claim 1 , wherein the machine learning model is trained by applying an oversampling technique to a plurality of login event inputs associated with authenticated user logins.
10. The method of claim 1 , wherein the machine learning model is trained with one or more of: a dataset of biometric data associated with the user logging into the account using a same credential, a dataset of biometric data associated with the user logging into the account using a different credential,
a dataset of biometric data associated with other users logging into the account using the same credential, or a dataset of biometric data associated with other users logging into their accounts.
11. The method of claim 1 , wherein the authentication of the user for the account comprises initiating a process of secondary authentication for the user.
12. A tangible, non-transitory computer-readable medium storing instructions that, when executed, cause a computer system to:
receive a login event input from a user, the login event input being associated with a session of the user logging into an account;
access a machine learning model;
authenticate, with the machine learning model, the user for the account, based at least in part on the login event input;
the login event input comprises one or more items of biometric data associated with the user;
an item of the one or more items of biometric data associated with the user is generated by interaction of the user with an input device for logging into the account, the interaction communicating a login credential of the user;
an item of the one or more items of biometric data associated with the user is keyboard event related biometric data, or mouse event related biometric data;
the machine learning model is trained by a plurality of login event inputs associated with sessions of the user logging into the account;
the machine learning model has been trained by a predetermined threshold number of sessions of the user logging into the account; and
a vector of mouse features is fused with a vector of keyboard features to generate a vector of login features.
13. The computer-readable medium of claim 12 , wherein the login event input comprises mouse movement features associated with the user performing a login into the account.
14. A system comprising:
a processor; and
memory in communication with the processor and storing instructions that,
when executed by the processor, cause the system to:
receive a login event input from a user, the login event input being
associated with a session of the user logging into an account;
access a machine learning model; authenticate, with the machine learning model, the user for the account, based at least in part on the login event input;
the login event input comprises one or more items of biometric data associated with the user;
an item of the one or more items of biometric data associated with the user is generated by interaction of the user with an input device for logging into the account, the interaction communicating a login credential of the user;
an item of the one or more items of biometric data associated with the user is keyboard event related biometric data, or mouse event related biometric data;
the machine learning model is trained by a plurality of login event inputs associated with sessions of the user logging into the account;
the machine learning model has been trained by a predetermined threshold number of sessions of the user logging into the account; and
a vector of mouse features is fused with a vector of keyboard features to generate a vector of login features.