-
Notifications
You must be signed in to change notification settings - Fork 523
New issue
Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.
By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.
Already on GitHub? Sign in to your account
Protections for AWS Bedrock #4270
Conversation
Rule: New - GuidelinesThese guidelines serve as a reminder set of considerations when proposing a new rule. Documentation and Context
Rule Metadata Checks
New BBR Rules
Testing and Validation
|
Does it make sense to have individual rules? Conceptually, it appears that the overall threat seems to be relatively similar and can be bucketed. |
This reverts commit c116468.
@Mikaayenson Grouped all content filter violation codes into one categroy. The remaining ones are all individual category by itself! |
This reverts commit 3c5f805.
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
Great work! - Maybe we can consolidate these rules to one with different alerts like Mika mentioned.
As discussed, we may want to take into consideration, from an analyst perspective, if the context of alerts from these rules indicate an active attack or if IR can assume breach if escalated. Some of them seem more "misconduct" based, then a signal of an attack.
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
Personally I think we should stick to cybersecurity, and avoid putting detections on topics like hate and insults. Additionally, I would suggest to bump all of these down to medium severity.
rules/integrations/aws_bedrock/aws_bedrock_high_confidence_misconduct_blocks_detected.toml
Show resolved
Hide resolved
rules/integrations/aws_bedrock/aws_bedrock_high_confidence_misconduct_blocks_detected.toml
Show resolved
Hide resolved
rules/integrations/aws_bedrock/aws_bedrock_multiple_word_policy_blocks_detected.toml
Show resolved
Hide resolved
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
Would suggest downing all of the risk scores and severities to medium.
(cherry picked from commit 801efb3)
(cherry picked from commit 801efb3)
(cherry picked from commit 801efb3)
(cherry picked from commit 801efb3)
Pull Request
Issue link(s): Protections for AWS Bedrock ( Onweek Research) https://github.com/elastic/security-team/issues/11054
As part of the ongoing effort to explore LLM guard tool and its applications to AWS bedrock models, this PR identifies a detection rule to detect anomalies or a hunting query for broader investigation(s) and security monitoring.
Summary - What I changed
New rule to detect unusual high denied topic block
New rule to detect unusual high confidence content Filter block
New rule to detect Sensitive information block
New rule to detect Word Policy block
New rule to detect Model Invocations without Guardrail Configuration
How To Test
Unusual High Denied Topic Block Test Details
ES|QL Query Execution
Sample Json
Unusual High Confidence Conetnt Block Test Details
ES|QL Query Execution
Sample Json
Unusual High Confidence Sensitive Information block Test Details
ES|QL Execution Summary
Sample JSON
Unusual High Word Policy Blocks Test Details
ES|QL Execution
Sample JSON
Multiple Model Invocations Without Guardrails Test Details
ES|QL Execution
Sample JSON