Classification of transport layer security certificates using artificial neural networks
Technologies for classification of web security certificates using artificial neural networks. Some of the example technologies disclosed herein are directed specifically at classification of TLS certificates using artificial neural networks. The technologies include methods for identifying malicious use and generation of web security certificates, by using deep neural networks. In one example embodiment, content of TLS certificates can be used as input for deep neural networks to successfully identify certificates of malicious actors as well as malicious patterns used by attackers.
1. A method, comprising:
receiving, in an artificial neural network (ANN), a plurality of groups of encoded parameters generated from a transport layer security (TLS) certificate, wherein the plurality of groups of encoded parameters comprises a first group of encoded parameters and a second group of encoded parameters, wherein the ANN comprises a first portion and a second portion separate from the first portion in processing inputs;
processing, by the first portion of the ANN, the first group of encoded para meters to generate first outputs for the TLS certificate, wherein the first group of encoded parameters are generated from a subject principal string of the TLS certificate;
processing, by the second portion of the ANN independent of the processing of the first portion, the second group of encoded parameters to generate second outputs for the TLS certificate, wherein the second group of encoded para meters are generated from an issuer principal string of the TLS certificate;
combining, in the ANN, the first outputs and the second outputs to generate a score; and
outputting, by the ANN, the score of the TLS certificate based on the plurality of groups of encoded parameters.
2. The method of claim 1 , wherein the ANN comprises a recurrent neural network (RNN) configured to process a first group of the plurality of groups of encoded parameters.
3. The method of claim 2 , wherein the RNN comprises a long short-term memory (LSTM) network.
4. The method of claim 1 , wherein the TLS certificate is a classified TLS certificate that has an issuer of the certificate and a subject of the certificate, the subject having a classification of being associated with a malicious activity, and wherein the method further comprises:
training the ANN to match the score with the classification of the subject classified TLS certificate.
5. The method of claim 1 , wherein the score is representative of the probability that the TLS certificate has a subject associated with a malicious activity selected from a group consisting of phishing and malware attack.
6. The method of claim 1 , further comprising:
comparing the score to a threshold value; and
determining that the TLS certificate is more likely than not associated with a malicious activity, based on the comparison of the score to the threshold value.
7. The method of claim 1 , further comprising:
processing the score to output a probability value related to the probability that the TLS certificate is associated with a malicious activity.
8. The method of claim 1 , wherein the subject principal string has a plurality of alphanumerical field values of the TLS certificate identifying attributes of the subject of the TLS certificate, and wherein the issuer principal string has a plurality of alphanumerical field values of the TLS certificate identifying attributes of the issuer of the TLS certificate.
9. The method of claim 8 , wherein the subject principal string comprises one or more fields selected from the group of fields consisting of an organization field, a company field, a state field, a location field, and a common name (CN) field, and wherein the issuer principal string comprises one or more fields selected from the group of fields consisting of an organization field, a company field, a state field, a location field, and a CN field.
10. The method of claim 1 , wherein the first portion of the ANN comprises a first long short-term memory (LSTM) network layer to process the first group of encoded parameters, and the second portion of the ANN comprises a second LSTM network layer to process the second group of encoded parameters.
11. The method of claim 10 , wherein the ANN comprises a third portion separate from the first portion and the second portion, the third portion includes a dense layer having at least one rectifier, and wherein the method further comprises:
receiving, in the dense layer, an encoded set of features of the TLS certificate, separate from the first group of encoded parameters and the second group of encoded parameters; and
outputting, by the dense layer, a dense layer output vector according to the encoded set of features.
12. The method of claim 11 , further comprising concatenating, in the ANN, the dense layer output vector, the first outputs of the first portion, the second outputs of the second portion to generate a concatenated vector.
13. The method of claim 12 , further comprising:
receiving, by a second dense layer of the ANN, the concatenated vector;
outputting, by the second dense layer, a second dense layer output vector according to the concatenated vector, wherein the second dense layer comprises a rectifier.
14. The method of claim 13 , further comprising:
receiving, by a dropout layer of the ANN, the second dense layer output vector;
generating, by the dropout layer, a reduced vector;
receiving, by a third dense layer of the ANN, the reduced vector, the third dense layer comprises a sigmoid function;
outputting, by the third dense layer, the score according to the reduced vector.
15. A non-transitory computer-readable storage medium comprising instructions that, when executed by a processing device, cause the processing device to:
input a plurality of groups of encoded parameters of a transport layer security (TLS) certificate into an artificial neural network (ANN), wherein the plurality of groups of encoded parameters comprises a first group of encoded para meters and a second group of encoded parameters, wherein the ANN comprises a first portion and a second portion separate from the first portion in processing inputs;
processing, by the first portion of the ANN, the first group of encoded parameters to generate first outputs for the TLS certificate, wherein the first group of encoded parameters are generated from a subject principal string of the TLS certificate;
processing, by the second portion of the ANN independent of the processing of the first portion, the second group of encoded parameters to generate second outputs for the TLS certificate, wherein the second group of encoded parameters are generated from an issuer principal string of the TLS certificate;
combining, in the ANN, the first outputs and the second outputs to generate a score; and
output the score according to an algorithm representative of the ANN and the plurality of encoded parameters.
16. The non-transitory computer-readable storage medium of claim 15 , wherein the ANN comprises a long short-term memory (LSTM) network.
17. The non-transitory computer-readable storage medium of claim 15 , wherein the TLS certificate is a classified TLS certificate that has an issuer of the certificate and a subject of the certificate, the subject having a classification of being associated with a malicious activity, and wherein the method further comprises:
training the ANN to match the score with the classification of the subject classified TLS certificate.
18. A computer system, comprising:
a processing device; and
memory in communication with the processing device and storing instructions that, when executed by the processing device, cause the processing device to:
input a plurality of groups of encoded para meters of a transport layer security (TLS) certificate into an artificial neural network (ANN), wherein the plurality of groups of encoded parameters comprises a first group of encoded para meters and a second group of encoded parameters, wherein the ANN comprises a first portion and a second portion separate from the first portion in processing inputs;
processing, by the first portion of the ANN, the first group of encoded para meters to generate first outputs for the TLS certificate, wherein the first group of encoded parameters are generated from a subject principal string of the TLS certificate;
processing, by the second portion of the ANN independent of the processing of the first portion, the second group of encoded para meters to generate second outputs for the TLS certificate, wherein the second group of encoded parameters are generated from an issuer principal string of the TLS certificate;
combining, in the ANN, the first outputs and the second outputs to generate a score; and
output the score according to an algorithm representative of the ANN and the plurality of encoded parameters.