Mobile application malicious behavior pattern detection method based on API call graph extraction and recording medium and device for performing the same

1. A mobile application malicious behavior pattern detection method based on Application Programming Interface (API) call graph extraction, the method comprising:

extracting an API Call Graph (ACG) representing a call flow of an API from benign applications and applications performing malicious behavior;

generating and vectorizing a training dataset for deep learning using the extracted ACG;

generating a deep learning algorithm prediction model by training with the vectorized training dataset;

extracting ACG features used in the malicious behavior from the generated prediction model and extracting a malicious behavior pattern from an intersection of the malicious applications; and

classifying a target application performing malicious behavior through a similarity comparison between the extracted malicious behavior pattern and a pattern extracted from the target application,

wherein the extracting the malicious behavior pattern comprises:

extracting the ACG features used in the malicious behavior using a deep learning visualization technique from the prediction model; and

grouping malicious applications having similar ACGs among the extracted ACG features and calculating the intersection of the malicious applications to extract the malicious behavior pattern.

2. The method of claim 1 , further comprising:

crawling and extracting framework API information from an Android homepage; and storing the extracted framework API information.

3. The method of claim 1 , wherein the extracting the

API Call Graph (ACG) comprises extracting the API Call Graph (ACG) representing the call flow using taint analysis.

4. The method of claim 1 , wherein the generating and vectorizing the training dataset comprises:

generating the training dataset using all the extracted ACGs; and

vectorizing the generated training dataset for deep learning.

5. The method of claim 1 , wherein the classifying the target application comprises listing the extracted malicious behavior patterns and measuring a Jaccard similarity between the malicious behavior pattern list and the ACG extracted from the target application.

6. A non-transitory computer-readable storage medium having recorded thereon a computer program for performing a mobile application malicious behavior pattern detection method based on Application Programming Interface (API) call graph extraction, wherein the method comprises:

extracting an API Call Graph (ACG) representing a call flow of an API from benign applications and applications performing malicious behavior;

generating and vectorizing a training dataset for deep learning using the extracted ACG;

generating a deep learning algorithm prediction model by training with the vectorized training dataset;

extracting ACG features used in the malicious behavior from the generated prediction model and extracting a malicious behavior pattern from an intersection of the malicious applications; and

classifying a target application performing malicious behavior through a similarity comparison between the extracted malicious behavior pattern and a pattern extracted from the target application,

wherein the extracting the malicious behavior pattern comprises:

extracting the ACG features used in the malicious behavior using a deep learning visualization technique from the prediction model; and

grouping malicious applications having similar ACGs among the extracted ACG features and calculating the intersection of the malicious applications to extract the malicious behavior pattern.

7. A mobile application malicious behavior pattern detection device based on Application Programming Interface (API) call graph extraction, the device comprising:

an API call graph extractor extracting an API Call Graph (ACG) representing a call flow of an API from benign applications and applications performing malicious behavior;

a training dataset vectorizer generating and vectorizing a training dataset for deep learning using the extracted ACG;

a machine learner generating a deep learning algorithm prediction model by training with the vectorized training dataset;

a pattern extractor extracting ACG features used in the malicious behavior from the generated prediction model and extract a malicious behavior pattern from an intersection of the malicious applications; and

a risk application classifier classifying a target application performing malicious behavior through a similarity comparison between the extracted malicious behavior pattern and a pattern extracted from the target application,

wherein the pattern extractor comprises:

an API call graph feature extractor extracting the ACG features used in the malicious behavior using a deep learning visualization technique from the prediction model; and

a malicious behavior pattern extractor grouping malicious applications having similar ACGs among the extracted ACG features and calculating the intersection of the malicious applications to extract the malicious behavior pattern.

8. The device of claim 7 , further comprising:

a framework API crawler extracting framework API information from an Android homepage; and

a framework API database to store the extracted framework API information.

9. The device of claim 7 , wherein the API call graph extractor extracts the API Call Graph (ACG) representing the call flow using taint analysis.

10. The device of claim 7 , wherein the training dataset vectorizer comprises:

a training dataset generator generating the training dataset using all the extracted ACGs; and

an embedding vectorizer vectorizing the generated training dataset for deep learning.

11. The device of claim 7 , wherein the risk application classifier comprises a Jaccard similarity comparator listing the extracted malicious behavior patterns and measuring a Jaccard similarity between the malicious behavior pattern list and the ACG extracted from the target application.