Method, system and server for self-healing of electronic apparatus

• US 10,075,477 B2
• Filed: 10/28/2015
• Issued: 09/11/2018
• Est. Priority Date: 06/02/2015
• Status: Active Grant

• Current Assignee: Wistron Corporation
• Original Assignee: Wistron Corporation
• Inventors: Chen, Chih-Ming
• Primary Examiner(s): Okeke, Izunna
• Assistant Examiner(s): Song, Hee K
• Application Number:
  US14/924,722
• Publication Number:
  US 20160359923A1
• Time in Patent Office: 1,049 Days
• CPC Class Codes: G06N 20/00 Machine learning
• CPC Class Codes: G06N 5/04 Inference or reasoning models
• CPC Class Codes: H04L 63/1416 Event detection, e.g. attack signature detection
• CPC Class Codes: H04L 63/205 involving negotiation or determination of the one or more network security mechanisms to be used, e.g. by negotiation between the client and the server or between peers or by selection according to the capabilities of the entities involved negotiation of communication capabilities H04L69/24

Abstract

A method, a system and a server for self-healing of an electronic apparatus are provided. In the method, the components installed in the electronic apparatus are sequentially called and checked to update each component. Then, a self-diagnosis is executed on the electronic apparatus to produce a diagnosis result and the diagnosis result is sent to a server. The server analyzes the diagnosis result, accordingly sets at least one policy for healing the electronic apparatus and sends the at least one policy back to the electronic apparatus. The electronic apparatus transforms the policy into at least one rule adapted for itself and performs the self-healing according to the rules.

Claims

1. A self-healing method of electronic apparatus, adapted to execute self-healing when at least one component in an electronic apparatus is updated, the self-healing method comprising:

sequentially calling and checking the components to update each of the components;

executing a self-diagnosis on the updated electronic apparatus to produce a diagnosis result, and sending the diagnosis result to a server;

analyzing the diagnosis result by the server to accordingly set at least one policy for healing the electronic apparatus, and sending the at least one policy back to the electronic apparatus comprising;

executing an anomaly detection on the diagnosis result by the server to find a plurality of outliers;

extracting a plurality of features in the outliers to generate a feature vector; and

taking the feature vector of each set of at least one sample as an input of a joint feature function to execute a structured prediction algorithm to predict the at least one policy used for healing the electronic apparatus; and

transforming the at least one policy into at least one rule adapted for the electronic apparatus, and performing the self-healing according to the rules.

2. The self-healing method as claimed in claim 1, wherein the step of sequentially calling and checking the components to update each of the components comprises:

obtaining a clone of the components installed in the electronic apparatus;

sequentially calling the components in the clone to check whether each of the components has an update;

pushing a context of the component to a stack when the component has the update, executing the update to the component, and extracting the context from the stack and modifying the context;

performing a simulation on the modified context and the update by using the clone to confirm a sanity of the context and the update;

updating the component corresponding to the context and the update in the electronic apparatus by using the clone if the context and the update have the sanity; and

calling a next component for updating until update of all of the components is completed.

3. The self-healing method as claimed in claim 2, wherein the step of executing the update to the component comprises:

modifying a file header, a program header, a relative offset of file content or an absolute offset of program content of the component.

4. The self-healing method as claimed in claim 2, wherein the step of sequentially calling the components in the clone comprises:

sorting the components according to at least one parameter of each of the components, so as to sequentially call the components, wherein the parameter comprises a risk level, a severity or a priority.

5. The self-healing method as claimed in claim 2, wherein the step of extracting the context from the stack and modifying the context comprises:

comparing the context before and after the modification, and transforming one of at least one register, heap variables, a driver interface, library dependencies, environment variables or a combination thereof of the modified context.

6. The self-healing method as claimed in claim 1, wherein the step of taking the feature vector of each set of the at least one sample as the input of the joint feature function to execute the structured prediction algorithm to predict the at least one policy used for healing the electronic apparatus further comprises:

taking the feature vector of each set of the at least one sample and at least one label as the input of the joint feature function to execute the structured prediction algorithm to predict the at least one policy used for healing the electronic apparatus.

7. The self-healing method as claimed in claim 6, wherein the step of taking the feature vector of each set of the at least one sample and the at least one label as the input of the joint feature function to execute the structured prediction algorithm to predict the at least one policy used for healing the electronic apparatus comprises:

establishing a machine learning model by using a semi-supervised learning method;

inputting the feature vector of the sample and the label to the machine learning model to obtain a plurality of feature vectors of predicted outputs;

combining the input feature vector and the obtained feature vectors of predicted outputs to generate the joint feature function; and

feeding back the joint feature function to the machine learning model to train the machine learning model, and accordingly predicting the at least one policy used for healing the electronic apparatus.

8. The self-healing method as claimed in claim 1, wherein the step of transforming the at least one policy into the rules adapted for the electronic apparatus and performing the self-healing according to the rules comprises:

merging the rules with a function of the corresponding component, covering the function of the corresponding component with the rules, or adding the rules after removing the function of the corresponding component.

9. A self-healing system of electronic apparatus, comprising:

an electronic apparatus, comprising;

a first communication interface, configured to connect a network;

a first processor, configured to;

sequentially call at least one component installed in the electronic apparatus to update each of the components; and

execute a self-diagnosis on the updated electronic apparatus to produce a diagnosis result, and transmit the diagnosis result through the first communication interface; and

a server, comprising;

a second communication interface, configured to connect the network and receive the diagnosis result transmitted by the first communication interface through the network; and

a second processor, configured to analyze the diagnosis result, and accordingly set at least one policy for healing the electronic apparatus and send the at least one policy back to the electronic apparatus through the second communication interface, wherein the second processor executes an anomaly detection on the diagnosis result to find a plurality of outliers, extracts a plurality of features in the outliers to generate a feature vector, and takes the feature vector of each set of at least one sample as an input of a joint feature function to execute a structured prediction algorithm to predict the at least one policy used for healing the electronic apparatus,

wherein the first processor of the electronic apparatus transforms the at least one policy into at least one rule adapted for the electronic apparatus and accordingly performs the self-healing of the electronic apparatus.

10. The self-healing system as claimed in claim 9, wherein the first processor obtains a clone of the components installed in the electronic apparatus, sequentially calls the components in the clone to check whether each of the components has an update, pushes a context of the component having the update to a stack to update the component, extracts the context from the stack and modifies the context, and performs a simulation on the modified context and the update by using the clone to confirm a sanity of the context and the update, and if the context and the update have the sanity, the first processor updates the component corresponding to the context and the update in the electronic apparatus by using the clone, and continually calls a next component for updating until update of all of the components is completed.

11. The self-healing system as claimed in claim 10, wherein the first processor modifies a file header, a program header, a relative offset of file content or an absolute offset of program content of the component.

12. The self-healing system as claimed in claim 10, wherein the first processor sorts the components according to at least one parameter of each of the components, so as to sequentially call the components, wherein the parameter comprises a risk level, a severity or a priority.

13. The self-healing system as claimed in claim 10, wherein the first processor compares the context before and after the modification when the first processor extracts the context from the stack and modifies the context, and transforms one of at least one register, heap variables, a driver interface, library dependencies, environment variables or a combination thereof of the modified context.

14. The self-healing system as claimed in claim 9, wherein the second processor is further configured to:

take the feature vector of each set of the at least one sample and at least one label as the input of the joint feature function to execute the structured prediction algorithm to predict the at least one policy used for healing the electronic apparatus.

15. The self-healing system as claimed in claim 14, wherein the second processor establishes a machine learning model by using a semi-supervised learning method, inputs the feature vector of the sample and the label to the machine learning model to obtain a plurality of feature vectors of predicted outputs, combines the input feature vector and the obtained feature vectors of predicted outputs to generate the joint feature function, and feeds back the joint feature function to the machine learning model to train the machine learning model, and accordingly predicts the at least one policy used for healing the electronic apparatus.

16. The self-healing system as claimed in claim 9, wherein the first processor merges the rules with a function of the corresponding component, covers the function of the corresponding component with the rules, or adds the rules after removing the function of the corresponding component.

17. A server, comprising:

a communication interface circuit, configured to connect a network and receive a diagnosis result transmitted by an electronic apparatus through the network, wherein the diagnosis result is produced by the electronic apparatus executing a self-diagnosis after the electronic apparatus executes an update to at least one component installed in the electronic apparatus; and

a processor, configured to analyze the diagnosis result, accordingly set at least one policy for healing the electronic apparatus, and send the at least one policy back to the electronic apparatus through the communication interface circuit to enable the electronic apparatus to perform a self-healing according to the at least one policy, wherein the processor executes an anomaly detection on the diagnosis result to find a plurality of outliers, extracts a plurality of features in the outliers to generate a feature vector, and takes the feature vector of each set of at least one sample as an input of a joint feature function to execute a structured prediction algorithm to predict the at least one policy used for healing the electronic apparatus.

18. The server as claimed in claim 17, wherein the processor is further configured to:

take the feature vector of each set of the at least one sample and at least one label as the input of the joint feature function to execute the structured prediction algorithm to predict the at least one policy used for healing the electronic apparatus.

19. The server as claimed in claim 18, wherein the processor establishes a machine learning model by using a semi-supervised learning method, inputs the feature vector of the sample and the label to the machine learning model to obtain a plurality of feature vectors of predicted outputs, combines the input feature vector and the obtained feature vectors of predicted outputs to generate the joint feature function, and feeds back the joint feature function to the machine learning model to train the machine learning model, and accordingly predicts the at least one policy used for healing the electronic apparatus.