Adaptive network attack prediction system
Machine-learning (ML) techniques and models are described for predicting the number and severity of network attacks within a specified timeframe, such as the next fifteen minutes. In some embodiments, the techniques including training a ML model based on features extracted from a training dataset and applying the trained ML model to estimate (a) the probability of an attack happening on an account within a specified timeframe; (b) how many attacks are predicted to occur in the specified timeframe (if any); and/or (c) the severity of the attacks predicted to occur. A system may deploy preventative measures based on the ML model output to counter or mitigate the effects of predicted and coordinated network attacks.
1. One or more non-transitory computer-readable media storing instructions, which, when executed by one or more hardware processors, cause:
identifying varying values for a set of features within historical activity associated with a user account for accessing one or more network services;
training, based on the varying values for the set of features, a machine-learning model to predict upcoming network attacks;
detecting activity associated with the user account;
generating, by the machine-learning model based on the detected activity, an output that identifies how many network attacks are predicted to occur up to a future horizon; and
executing, as a function of the output of the machine learning model, a rule that compares how many network attacks are predicted to occur up to the future horizon to a configurable threshold number, wherein executing the rule performs at least one operation to secure the user account responsive, at least in part, to determining that a number of network attacks predicted to occur up to the future horizon exceeds the configurable threshold number.
2. The one or more non-transitory computer-readable media of claim 1 , wherein the set of varying values include outlier scores for one or more features; wherein the outlier score is determined based at least in part on a comparison of an observed value with a behavioral profile associated with the user account.
3. The one or more non-transitory computer-readable media of claim 2 , wherein outlier scores are determined for at least one of a location, network address, browser, language, or network provider associated with detected user activity.
4. The one or more non-transitory computer-readable media of claim 1 , wherein the set of varying values include aggregated values for one or more aggregated features; wherein the aggregated values track one or more actions within a threshold timeframe.
5. The one or more non-transitory computer-readable media of claim 4 , wherein the one or more aggregated features include at least one of a number of vulnerability scanners, directory traversals, standard query language (SQL) injection attempts, login successes, login failures, blocked addresses, cities in login failures, or countries in login failures within the threshold timeframe.
6. The one or more non-transitory computer-readable media of claim 1 , wherein the machine-learning model includes one or more decision trees; wherein training the machine-learning model comprises splitting training examples from the historical activity based at least in part on how predictive the varying feature values are of network attacks.
7. The one or more non-transitory computer-readable media of claim 6 , wherein the instructions further cause: pruning the one or more decision trees based at least in part on how predictive the varying feature values are of network attacks.
8. The one or more non-transitory computer-readable media of claim 1 , wherein the instructions further cause: adjusting at least one model hyperparameter to balance between a precision and a recall of the machine-learning model.
9. The one or more non-transitory computer-readable media of claim 1 , wherein the instructions further cause: selecting a set of features to train the machine-learning model based on measure of the machine-learning model precision and balance for varying sets of features.
10. The one or more non-transitory computer-readable media of claim 1 , wherein detecting the activity associated with the user account comprises detecting one or more login attempts.
11. The one or more non-transitory computer-readable media of claim 1 , wherein generating the prediction comprises traversing one or more decision trees based on values for the set of features extracted from the detected activity.
12. The one or more non-transitory computer-readable media of claim 1 , wherein the output further includes a predicted probability associated with how many network attacks are predicted up to the future horizon.
13. The one or more non-transitory computer-readable media of claim 1 , wherein the at least one operation includes at least one of locking the user account, sending a user a one-time password, or enabling two-factor authentication.
14. The one or more non-transitory computer-readable media of claim 1 , wherein the output identifies a predicted attack on the user account within a specified timeframe.
15. The one or more non-transitory computer-readable media of claim 1 , wherein the output identifies a predicted attack and severity based on log events associated with online shoppers.
16. The one or more non-transitory computer-readable media of claim 1 , wherein the set of features include outlier scores for a location, network address, browser, language, and network provider; wherein the set of features further includes a number of vulnerability scanners, directory traversals, standard query language (SQL) injection attempts, login successes, login failures, blocked addresses, cities in login failures, and countries in login failures within a threshold timeframe.
17. A system comprising:
one or more hardware processors;
one or more non-transitory computer-readable media storing instructions, which, when executed by one or more hardware processors, cause performance of operations comprising:
identifying varying values for a set of features within historical activity associated with a user account for accessing one or more network services;
training, based on the varying values for the set of features, a machine-learning model to predict upcoming network attacks;
detecting activity associated with the user account;
generating, by the machine-learning model based on the detected activity, an output that identifies how many network attacks are predicted to occur up to a future horizon; and
executing, as a function of the output of the machine learning model, a rule that compares how many network attacks are predicted to occur up to the future horizon to a configurable threshold number, wherein executing the rule performs at least one operation to secure the user account responsive, at least in part, to determining that a number of network attacks predicted to occur up to the future horizon exceeds the configurable threshold number.
18. A method comprising:
identifying varying values for a set of features within historical activity associated with a user account for accessing one or more network services;
training, based on the varying values for the set of features, a machine-learning model to predict upcoming network attacks;
detecting activity associated with the user account; and
generating, by the machine-learning model based on the detected activity, an output that identifies how many network attacks are predicted to occur up to a future horizon; and
executing, as a function of the output of the machine learning model, a rule that compares how many network attacks are predicted to occur up to the future horizon to a configurable threshold number, wherein executing the rule performs at least one operation to secure the user account responsive, at least in part, to determining that a number of network attacks predicted to occur up to the future horizon exceeds the configurable threshold number.
19. The one or more non-transitory computer-readable media of claim 12 , wherein the rule further compares the predicted probability to a threshold probability, wherein the at least one operation is performed only if the predicted probability satisfies the threshold probability.
20. The one or more non-transitory computer-readable media of claim 1 , wherein the configurable threshold number is greater than zero.