Workload pattern detection
A computer-implemented method, computer program product and computer system are provided. A processor retrieves instructions for a plurality of processes. A processor generates a sample of the instructions for the plurality of processes. A processor determines a probability that the sample includes instructions from malicious processes. A processor, in response to the probability exceeding a threshold value, restricts one or more aspects of operation for at least one process of the plurality of processes.
1. A computer-implemented method comprising:
retrieving instructions for a plurality of processes;
generating a sample of the instructions for the plurality of processes, wherein the sample of the instructions are non-sequential in regards to the instructions execution order, and wherein the sample of the instructions are randomly selected and include filtering aspects that attribute the sample to a device or process the sample of the instructions are retrieved from;
determining, using a machine learning model, a probability that the sample includes instructions from malicious processes;
in response to the probability exceeding a threshold value, restricting one or more aspects of operation for at least one process of the plurality of processes;
filtering, in response to the probability exceeding the threshold value, at least one process from the sample based on one or more identifiers corresponding to the filtering aspects that attribute the sample to the device or the process;
and generating a sample for at least each process and device that includes the instructions from the malicious process.
2. The computer-implemented method of claim 1 , wherein the machine learning model is a logistic regression model.
3. The computer-implemented method of claim 1 , wherein the machine learning model includes on or more of the following models: a deep neural network (DNN) or a support vector machine (SVM).
4. The computer-implemented method of claim 1 , the method further comprising:
in response to the probability exceeding the threshold value, filtering at least one process from the sample; and
determining a filtered probability that the filtered sample includes instructions from malicious processes.
5. The computer-implemented method of claim 1 , wherein the determining the probability that the sample includes instructions from malicious processes is performed by a hardware module connected to a processor executing the more than one process.
6. A computer program product comprising:
one or more computer-readable storage media and program instructions stored on the one or more computer-readable storage media, the program instructions comprising:
program instructions to retrieve instructions for a plurality of processes;
program instructions to generate a sample of the instructions for the plurality of processes, wherein the sample of the instructions are non-sequential in regards to the instructions execution order, and wherein the sample of the instructions are randomly selected and include filtering aspects that attribute the sample to a device or process the sample of the instructions are retrieved from;
program instructions to determine, using a machine learning model, a probability that the sample includes instructions from malicious processes;
program instructions, in response to the probability exceeding a threshold value, to restrict one or more aspects of operation for at least one process of the plurality of processes;
program instructions to filter, in response to the probability exceeding the threshold value, at least one process from the sample based on one or more identifiers corresponding to the filtering aspects that attribute the sample to the device or the process;
and program instructions to generate a sample for at least each process and device that includes the instructions from the malicious process.
7. The computer program product of claim 6 , wherein the machine learning model is a logistic regression model.
8. The computer program product of claim 6 , wherein the machine learning model includes on or more of the following models: a deep neural network (DNN) or a support vector machine (SVM).
9. The computer program product of claim 7 , the program instructions further comprising:
program instructions, in response to the probability exceeding the threshold value, to filter at least one process from the sample; and
program instructions to determine a filtered probability that the filtered sample includes instructions from malicious processes.
10. The computer program product of claim 6 , wherein the determining the probability that the sample includes instructions from malicious processes is performed by a hardware module connected to a processor executing the more than one process.
11. A computer system comprising:
one or more computer processors;
one or more computer readable storage media;
and program instructions stored on the computer readable storage media for execution by at least one of the one or more processors, the program instructions comprising:
program instructions to retrieve instructions for a plurality of processes;
program instructions to generate a sample of the instructions for the plurality of processes, wherein the sample of the instructions are non-sequential in regards to the instructions execution order, and wherein the sample of the instructions are randomly selected and include filtering aspects that attribute the sample to a device or process the sample of the instructions are retrieved from;
program instructions to determine, using a machine learning model, a probability that the sample includes instructions from malicious processes;
program instructions, in response to the probability exceeding a threshold value, to restrict one or more aspects of operation for at least one process of the plurality of processes;
program instructions to filter, in response to the probability exceeding the threshold value, at least one process from the sample based on one or more identifiers corresponding to the filtering aspects that attribute the sample to the device or the process;
and program instructions to generate a sample for at least each process and device that includes the instructions from the malicious process.
12. The computer system of claim 11 , the program instructions further comprising:
program instructions, in response to the probability exceeding the threshold value, to filter at least one process from the sample; and
program instructions to determine a filtered probability that the filtered sample includes instructions from malicious processes.
13. The computer-implemented method of claim 1 , wherein determining the probability that the sample includes instructions from the malicious process further comprises:
training the machine learning model based on a training data set, wherein the training data set is built from samples of the instructions for the plurality of processes retrieved, wherein the samples of the instructions retrieved include pre-classifications as either a malicious or a non-malicious program;
determining, using the machine learning model, a probability for each of the samples from the training data set include instructions from the malicious process, wherein the machine learning model uses logistic regression in determining the probability for each of the samples from the training data set;
verifying, the probability generated by the machine learning model for each of the samples corresponds to the pre-classifications; and
deploying the machine learning model for malicious code detection.
14. The computer-implemented method of claim 1 , wherein the instructions retrieved for the plurality of processes are binary operational codes executed by the plurality of processes, and wherein instruction data includes any addresses or registers accessed by the instructions corresponding to the plurality of processes.
15. The computer-implemented method of claim 1 , wherein determining the probability that the sample includes instructions from the malicious process further comprises:
creating statistical distribution data of the instructions;
observing, by the machine learning model, a distribution of potentially malicious instructions; and
increasing the probability that the sample includes instructions from malicious processes based on the observing of the distribution of the potentially malicious instructions.
16. The computer-implemented method of claim 1 , further comprising:
generating a subsample of the instructions, wherein the subsample filters out one or more processes from the sample with the probability exceeding the threshold value;
determining, using the machine learning model, that the probability the subsample includes instructions from the malicious process is less than the probability that the sample includes instructions from the malicious process;
determining that the one or more processes filtered from the subsample contributed to a maliciousness of the sample; and
adding the one or more processes back to the sample and filtering out one or more different processes for a next subsample.