Automatic evalution of remediation plans using exploitability risk modeling
Systems and methods are provided to build a machine learned exploitability risk model that predicts, based on the characteristics of a set of machines, a normalized risk score quantifying the risk that the machines are exploitable by a set of attacks. To build the model, a training dataset is constructed by labeling characteristic data of a population of machines with exploitation test results obtained by simulating a set of attacks on the population. The model is trained using the training data to accurately predict a probability that a given set of machines is exploitable by the set of attacks. In embodiments, the model may be used to make quick assessments about how vulnerable a set of machines are to the set of attacks. In embodiments, the model may be used to compare the effectiveness of different remediation actions to protect against the set of attacks.
1. A system, comprising:
one or more hardware processors with associated memory that implement a remediation planning system, configured to:
implement an exploitability risk model trained according to a machine learning technique, wherein the exploitability risk model is trained to determine, based on characteristics data of an individual set of one or more machines, a normalized risk score of the individual set of machines for a set of attacks;
determine, using the exploitability risk model and based on characteristics data of a set of machines, a risk score for the set of machines;
determine a plurality of remediation actions for the set of machines to reduce the risk score;
generate a plurality of candidate remediation plans that include different permutations of the remediation actions;
apply the candidate remediation plans to the characteristics data to generate updated instances of the characteristics data, wherein the application simulates performance of the candidate remediation plans on the set of machines without performing the candidate remediation plans on the set of machines;
determine, using the exploitability risk model and based on the updated instances of characteristics data, respective reductions of the risk score caused by the candidate remediation plans; and
provide an output indicating at least one of the candidate remediation plans as a selected remediation plan and a ranking of at least some of the candidate remediation plans, wherein the selected remediation plan and the ranking are determined based at least in part on the respective reductions of the risk score caused by the candidate remediation plans.
2. The system of claim 1 , wherein
the remediation planning system is implemented as part of a network-accessible service, configured to:
collect the characteristics data from the machines of the client; and
provide the selected remediation plan as a recommendation via a graphical user interface of the network-accessible service.
3. The system of claim 2 , wherein the network-accessible service is configured to:
monitor the risk score of the machines of the client over time; and
provide the recommendation in response to a detection that the risk score exceeds a risk threshold.
4. The system of claim 1 ,
wherein the ranking is determined based at least in part on the respective costs of the candidate remediation plans.
5. The system of claim 1 , wherein the exploitability risk model is a logistic regression model and trained to output a probability that the set of machines is exploitable by at least one of the set of attacks.
6. The system of claim 1 , wherein the remediation planning system is configured to determine the selected remediation plan to include one or more of:
applying a software patch to the set of machines;
disabling a feature provided by the set of machines;
removing one or more access privileges of one or more users from the set of machines;
updating one or more configuration settings of the set of machines; or
isolating a particular machine in the set of machines.
7. The system of claim 1 , wherein the remediation planning system is configured to retrieve the plurality of remediation actions from a library of remediation actions, wherein the library associates different types of remediation actions with different types of attacks.
8. A method comprising:
performing, by one or more hardware processors with associated memory that implement a remediation planning system:
implementing an exploitability risk model trained according to a machine learning technique, wherein the machine learning model is trained to determine, based on characteristics data of an individual set of one or more machines, a normalized risk score of the individual set of machines for a set of attacks;
determining, using the exploitability risk model and based on characteristics data of a set of machines, a risk score for the set of machines;
determining a plurality of remediation actions for the set of machines to reduce the risk score;
generating a plurality of candidate remediation plans that include different permutations of the remediation actions;
applying the candidate remediation plans to the characteristics data to generate updated instances of the characteristics data, wherein the application simulates performance of the candidate remediation plans on the set of machines without performing the candidate remediation plans on the set of machines;
determining, using the exploitability risk model and based on the updated instances of characteristics data, respective reductions of the risk score caused by the candidate remediation plans; and
providing an output indicating at least one of the candidate remediation plans as a selected remediation plan and a ranking of at least some of the candidate remediation plans, wherein the selected remediation plan and the ranking are determined based at least in part on the respective reductions of the risk score caused by the candidate remediation plans.
9. The method of claim 8 ,
wherein the remediation planning system is implemented as part of a network-accessible service; and
further comprising performing, by the network-accessible service:
collecting the characteristics data from the machines of the client; and
providing the selected remediation plan as a recommendation via a graphical user interface (GUI) of the network-accessible service.
10. The method of claim 9 , wherein the network-accessible service is configured to:
monitoring the risk score of the machines of the client over time; and
providing the recommendation in response to a determination that the risk score exceeds a risk threshold.
11. The method of claim 9 , further comprising performing, by the network-accessible service:
receiving, via the GUI, user input to modify the selected remediation plan;
generating, in accordance with the user input, a workflow to perform the selected remediation plan; and
providing the workflow via the GUI.
12. The method of claim 8 ,
wherein the ranking is determined based at least in part on the respective costs of the candidate remediation plans.
13. The method of claim 12 ,
wherein the respective costs are based on different cost metrics including a number of man hours associated with individual remediation plans and a amount of downtime associated with individual remediation plans.
14. The method of claim 13 , further comprising performing, by the remediation planning system:
storing, in a library of remediation actions, respective cost metrics of individual remediation actions; and
determining the respective costs of the selected remediation plans based at least in part on the cost metrics of respective remediation actions in the selected remediation plans.
15. The method of claim 8 , wherein the exploitability risk model is a logistic regression model and trained to output a probability that the set of machines is exploitable by at least one of the set of attacks.
16. One or more non-transitory computer-accessible storage media storing program instructions that when executed on or across one or more processors implement at least a portion of a remediation planning system and cause the remediation planning system to:
implement an exploitability risk model trained according to a machine learning technique, wherein the machine learning model is trained to determine, based on characteristics data of a an individual set of one or more machines, a normalized risk score of the individual set of machines for a set of attacks;
determine, using the exploitability risk model and based on characteristics data of a set of machines, a risk score for the set of machines;
determine a plurality of remediation actions for the set of machines to reduce the risk score;
generate a plurality of candidate remediation plans that include different permutations of the remediation actions;
apply the candidate remediation plans to the characteristics data to generate updated instances of the characteristics data, wherein the application simulates performance of the candidate remediation plans on the set of machines without performing the candidate remediation plans on the set of machines;
determine, using the exploitability risk model and based on the updated instances of characteristics data, respective reductions of the risk score caused by the candidate remediation plans; and
provide an output indicating at least one of the candidate remediation plans as a selected remediation plan and a ranking of at least some of the candidate remediation plans, wherein the selected remediation plan and the ranking are determined based at least in part on the respective reductions of the risk score caused by the candidate remediation plans.
17. The one or more non-transitory computer-accessible storage media of claim 16 , wherein the set of machines include one or more virtual machines.
18. The one or more non-transitory computer-accessible storage media of claim 16 , wherein the program instructions when executed on or across the one or more processors cause the remediation planning system to:
determine a rank metric for each of the candidate remediation plans based at least in part on the reduction of risk score caused by the candidate remediation plan and a respective cost of the candidate remediation plan; and
wherein the ranking of the candidate remediation plans is performed according to their respective rank metrics.
19. The one or more non-transitory computer-accessible storage media of claim 16 , wherein the program instructions when executed on or across the one or more processors cause the remediation planning system to:
monitor the risk score for the machines of the client over time; and
provide the output in response to a determination that the risk score meets a specified risk criterion.
20. The one or more non-transitory computer-accessible storage media of claim 16 , wherein the program instructions when executed on or across the one or more processors cause the remediation planning system to:
receive, via a graphical user interface (GUI), user input to modify the selected remediation plan;
generate, in accordance with the user input, a workflow to perform the selected remediation plan; and
provide the workflow via the GUI.