Multi-level classifier based access control

1. A computer implemented method of access control for a user device having at least one component for determining behaviors of a user of the user device, the method comprising:

accessing a first machine learning classifier trained based on at least one prior behavior of the user using the user device, the first machine learning classifier classifying user behavior as compliant or non-compliant such that compliant behavior is determined by the first machine learning classifier to be consistent with prior behavior for permitting access to the user device; and

responsive to a determination by the first machine learning classifier that a subsequent behavior is classified as non-compliant, accessing a second machine learning classifier trained based on at least one prior behavior of the user using the user device where the prior behavior is classified as non-compliant by the first machine learning classifier, the second machine learning classifier classifying user behavior as compliant or non-compliant such that compliant behavior is determined by the second machine learning classifier to be consistent with prior behavior for permitting access to the user device in the event of non-compliant classification by the first machine learning classifier, and further responsive to the determination:

responsive to a determination that the subsequent behavior is classified as non-compliant by the second machine learning classifier:

requesting a credential-based authentication of the user;

constructively training the second machine learning classifier based on the subsequent behavior and a result of the credential-based authentication by providing the subsequent behavior as a training example for the second machine learning classifier; and

permitting access to the user device in response to the credential-based authentication; and

responsive to a determination that the subsequent behavior is classified by the first machine learning classifier as compliant, constructively training the first machine learning classifier based on the subsequent behavior as a compliant behavior by providing the subsequent behavior as a training example for the first machine learning classifier.

2. The method of claim 1 , further comprising, responsive to a determination that the subsequent behavior is classified by the first machine learning classifier as non-compliant, constructively training the first machine learning classifier based on the subsequent behavior as a non-compliant behavior by providing the subsequent behavior as a training example for the first machine learning classifier.

3. The method of claim 1 , wherein the component is one or more of: a location sensor; a position sensor; an orientation sensor; an accelerometer; an input device; a touch-screen; a temperature sensor; a time determiner; a pressure sensor; an olfactory sensor; a chemical sensor; a biometric sensor; a heart rate sensor; a cardiogram generator; a sound sensor; a voice recognition component; a handwriting recognition component; a global positioning system; and a gyroscope.

4. The method of claim 1 , wherein the credential-based authentication includes one or more of: an authentication scheme using a user identifier and password; a key-based user authentication scheme; a token-based user authentication scheme; and a multi-factor authentication scheme in which authentication is requested via a different device.

5. A computer system comprising:

a processor; and

a memory storing computer program code for performing a method of access control for a user device having at least one component for determining behaviors of a user of the user device, the method comprising:

accessing a first machine learning classifier trained based on at least one prior behavior of the user using the user device, the first machine learning classifier classifying user behavior as compliant or non-compliant such that compliant behavior is determined by the first machine learning classifier to be consistent with prior behavior for permitting access to the user device; and

responsive to a determination by the first machine learning classifier that a subsequent behavior is classified as non-compliant, accessing a second machine learning classifier trained based on at least one prior behavior of the user using the user device where the prior behavior is classified as non-compliant by the first machine learning classifier, the second machine learning classifier classifying user behavior as compliant or non-compliant such that compliant behavior is determined by the second machine learning classifier to be consistent with prior behavior for permitting access to the user device in the event of non-compliant classification by the first machine learning classifier, and further responsive to the determination:

responsive to a determination that the subsequent behavior is classified as non-compliant by the second machine learning classifier:

requesting a credential-based authentication of the user;

constructively training the second machine learning classifier based on the subsequent behavior and a result of the credential-based authentication by providing the subsequent behavior as a training example for the second machine learning classifier; and

permitting access to the user device in response to the credential-based authentication; and

responsive to a determination that the subsequent behavior is classified by the first machine learning classifier as compliant, constructively training the first machine learning classifier based on the subsequent behavior as a compliant behavior by providing the subsequent behavior as a training example for the first machine learning classifier.

6. A non-transitory computer-readable storage medium comprising computer program code to, when loaded into a computer system and executed thereon, cause the computer system to perform a method of access control for a user device having at least one component for determining behaviors of a user of the user device, the method comprising:

accessing a first machine learning classifier trained based on at least one prior behavior of the user using the user device, the first machine learning classifier classifying user behavior as compliant or non-compliant such that compliant behavior is determined by the first machine learning classifier to be consistent with prior behavior for permitting access to the user device; and

responsive to a determination by the first machine learning classifier that a subsequent behavior is classified as non-compliant, accessing a second machine learning classifier trained based on at least one prior behavior of the user using the user device where the prior behavior is classified as non-compliant by the first machine learning classifier, the second machine learning classifier classifying user behavior as compliant or non-compliant such that compliant behavior is determined by the second machine learning classifier to be consistent with prior behavior for permitting access to the user device in the event of non-compliant classification by the first machine learning classifier, and further responsive to the determination:

responsive to a determination that the subsequent behavior is classified as non-compliant by the second machine learning classifier:

requesting a credential-based authentication of the user;

constructively training the second machine learning classifier based on the subsequent behavior and a result of the credential-based authentication by providing the subsequent behavior as a training example for the second machine learning classifier; and

permitting access to the user device in response to the credential-based authentication; and

responsive to a determination that the subsequent behavior is classified by the first machine learning classifier as compliant, constructively training the first machine learning classifier based on the subsequent behavior as a compliant behavior by providing the subsequent behavior as a training example for the first machine learning classifier.