```ThreatDeck detection pack (spl) version 2026.11.22-22e41752``` ```Generated 2026-11-22. 12 queries (8 ports of rules first written for another platform).``` ```Source: https://threatdeck.net/detections/ . Hunting queries are starting points: tune before alerting. Defensive use only.``` ```ATT&CK ids are MITRE ATT&CK Enterprise v20.1. Adapted rules keep their original licence (see attribution lines).``` ```Bedrock invocation probing: failed model calls across models or regions``` ```id: llmjacking-bedrock-invoke-validation-probe | maturity: hunt | techniques: LLM10, AML.T0034, AML.T0040, AML.T0012``` ```MITRE ATT&CK: T1078.004, T1496.004``` ```log source: AWS CloudTrail in Splunk (sourcetype aws:cloudtrail), eventSource bedrock.amazonaws.com``` ```page: https://threatdeck.net/detections/llmjacking-bedrock-invoke-validation-probe/``` ```source: https://sysdig.com/blog/llmjacking-stolen-cloud-credentials-used-in-new-ai-attack/``` ```source: https://permiso.io/blog/exploiting-hosted-models``` ```source: https://docs.aws.amazon.com/bedrock/latest/userguide/logging-using-cloudtrail.html``` index=aws sourcetype=aws:cloudtrail eventSource="bedrock.amazonaws.com" eventName IN ("InvokeModel", "InvokeModelWithResponseStream", "Converse", "ConverseStream") errorCode IN ("ValidationException", "AccessDenied", "AccessDeniedException") | eval principal=coalesce('userIdentity.arn', 'userIdentity.accessKeyId') | bin _time span=1h | stats count AS failed_calls dc(awsRegion) AS regions dc(requestParameters.modelId) AS models values(errorCode) AS errors values(userAgent) AS user_agents values(sourceIPAddress) AS source_ips BY _time principal | where regions >= 2 OR models >= 3 OR failed_calls >= 10 | sort - failed_calls ```Bedrock model-access enablement APIs called with a long-term access key (SPL port)``` ```id: llmjacking-bedrock-model-access-apis-from-access-key | maturity: alert | techniques: LLM10, AML.T0034, AML.T0012, AML.T0040``` ```MITRE ATT&CK: T1078.004, T1496.004``` ```log source: Splunk: AWS CloudTrail (Splunk Add-on for AWS, sourcetype aws:cloudtrail)``` ```note: Set index= to the index that holds your CloudTrail data.``` ```page: https://threatdeck.net/detections/llmjacking-bedrock-model-access-apis-from-access-key/``` ```source: https://permiso.io/blog/exploiting-hosted-models``` ```source: https://sysdig.com/blog/growing-dangers-of-llmjacking/``` ```source: https://docs.aws.amazon.com/bedrock/latest/userguide/logging-using-cloudtrail.html``` ```source: https://docs.aws.amazon.com/awscloudtrail/latest/userguide/query-limitations.html``` index=aws sourcetype=aws:cloudtrail eventSource="bedrock.amazonaws.com" eventName IN ("GetFoundationModelAvailability", "PutUseCaseForModelAccess", "GetUseCaseForModelAccess", "CreateFoundationModelAgreement", "PutFoundationModelEntitlement", "ListFoundationModelAgreementOffers") userIdentity.accessKeyId="AKIA*" | table _time eventName awsRegion sourceIPAddress userAgent errorCode userIdentity.arn userIdentity.accessKeyId recipientAccountId | sort - _time ```Bedrock model invocation logging configuration deleted (SPL port)``` ```id: llmjacking-bedrock-invocation-logging-deleted | maturity: alert | techniques: AML.T0138, LLM10, AML.T0034``` ```MITRE ATT&CK: T1685.002``` ```log source: Splunk: AWS CloudTrail (Splunk Add-on for AWS, sourcetype aws:cloudtrail)``` ```note: Set index= to the index that holds your CloudTrail data.``` ```page: https://threatdeck.net/detections/llmjacking-bedrock-invocation-logging-deleted/``` ```source: https://sysdig.com/blog/growing-dangers-of-llmjacking/``` ```source: https://docs.aws.amazon.com/bedrock/latest/userguide/model-invocation-logging.html``` ```source: https://docs.aws.amazon.com/bedrock/latest/userguide/logging-using-cloudtrail.html``` index=aws sourcetype=aws:cloudtrail eventSource="bedrock.amazonaws.com" eventName="DeleteModelInvocationLoggingConfiguration" | table _time userIdentity.arn userIdentity.accessKeyId sourceIPAddress userAgent awsRegion recipientAccountId errorCode ```Bedrock reconnaissance burst: model, availability and logging checks in one hour (SPL port)``` ```id: llmjacking-bedrock-recon-sequence | maturity: hunt | techniques: LLM10, AML.T0034, AML.T0040, AML.T0069``` ```MITRE ATT&CK: T1526, T1078.004``` ```log source: Splunk: AWS CloudTrail (Splunk Add-on for AWS, sourcetype aws:cloudtrail)``` ```note: Set index= to the index that holds your CloudTrail data.``` ```page: https://threatdeck.net/detections/llmjacking-bedrock-recon-sequence/``` ```source: https://sysdig.com/blog/llmjacking-stolen-cloud-credentials-used-in-new-ai-attack/``` ```source: https://sysdig.com/blog/growing-dangers-of-llmjacking/``` ```source: https://permiso.io/blog/exploiting-hosted-models``` ```source: https://docs.aws.amazon.com/awscloudtrail/latest/userguide/query-limitations.html``` index=aws sourcetype=aws:cloudtrail eventSource="bedrock.amazonaws.com" eventType="AwsApiCall" eventName IN ("ListFoundationModels", "GetFoundationModelAvailability", "ListFoundationModelAgreementOffers", "ListInferenceProfiles", "GetModelInvocationLoggingConfiguration", "GetUseCaseForModelAccess", "InvokeModel", "InvokeModelWithResponseStream", "Converse", "ConverseStream") earliest=-7d | eval is_check=if(in(eventName, "GetModelInvocationLoggingConfiguration", "GetFoundationModelAvailability", "ListFoundationModelAgreementOffers"), 1, 0) | bin _time span=1h | stats dc(eventName) AS distinct_apis values(eventName) AS apis count AS calls sum(is_check) AS checks BY _time userIdentity.arn sourceIPAddress | where distinct_apis >= 3 AND checks > 0 | sort - distinct_apis - calls ```Bedrock calls from a long-term access key with a browser, empty or scripting user agent (SPL port)``` ```id: llmjacking-bedrock-access-key-unusual-user-agent | maturity: hunt | techniques: LLM10, AML.T0034, AML.T0012``` ```MITRE ATT&CK: T1078.004, T1496.004``` ```log source: Splunk: AWS CloudTrail (Splunk Add-on for AWS, sourcetype aws:cloudtrail)``` ```note: Set index= to the index that holds your CloudTrail data.``` ```page: https://threatdeck.net/detections/llmjacking-bedrock-access-key-unusual-user-agent/``` ```source: https://permiso.io/blog/exploiting-hosted-models``` ```source: https://docs.aws.amazon.com/bedrock/latest/userguide/logging-using-cloudtrail.html``` index=aws sourcetype=aws:cloudtrail eventSource="bedrock.amazonaws.com" eventType="AwsApiCall" userIdentity.accessKeyId="AKIA*" earliest=-14d | fillnull value="" userAgent | where userAgent="" OR like(userAgent, "Mozilla/%") OR like(userAgent, "axios/%") OR like(userAgent, "%aiohttp%") OR like(userAgent, "python-urllib3/%") | stats count AS calls dc(eventName) AS distinct_apis values(sourceIPAddress) AS source_ips min(_time) AS first_seen max(_time) AS last_seen BY userIdentity.accessKeyId userIdentity.arn userAgent | convert ctime(first_seen) ctime(last_seen) | sort - calls ```Bedrock invocation volume far above a principal's own baseline (SPL port)``` ```id: llmjacking-bedrock-invocation-volume-spike | maturity: hunt | techniques: LLM10, AML.T0034, AML.T0040``` ```MITRE ATT&CK: T1496.004``` ```log source: Splunk: AWS CloudTrail (Splunk Add-on for AWS, sourcetype aws:cloudtrail)``` ```note: Set index= to the index that holds your CloudTrail data.``` ```page: https://threatdeck.net/detections/llmjacking-bedrock-invocation-volume-spike/``` ```source: https://sysdig.com/blog/growing-dangers-of-llmjacking/``` ```source: https://docs.aws.amazon.com/bedrock/latest/userguide/logging-using-cloudtrail.html``` ```source: https://docs.aws.amazon.com/awscloudtrail/latest/userguide/query-limitations.html``` index=aws sourcetype=aws:cloudtrail eventSource="bedrock.amazonaws.com" eventType="AwsApiCall" eventName IN ("InvokeModel", "InvokeModelWithResponseStream", "Converse", "ConverseStream") earliest=-14d@h | where isnull(errorCode) | eval principal='userIdentity.arn' | bin _time span=1h | stats count AS invocations dc(requestParameters.modelId) AS models dc(awsRegion) AS regions BY principal _time | eval in_window=if(_time >= relative_time(now(), "-1d@h"), 1, 0) | eval baseline_invocations=if(in_window=0, invocations, null()) | eventstats avg(baseline_invocations) AS baseline_avg count(baseline_invocations) AS baseline_hours BY principal | where in_window=1 AND invocations > 100 AND (isnull(baseline_avg) OR invocations > 5 * baseline_avg) | fields _time principal invocations models regions baseline_avg baseline_hours | sort - invocations ```One source IP using several AWS principals against Bedrock (SPL port)``` ```id: llmjacking-bedrock-many-keys-one-source-ip | maturity: hunt | techniques: AML.T0140, LLM10, AML.T0034, AML.T0012``` ```MITRE ATT&CK: T1496.004, T1078.004``` ```log source: Splunk: AWS CloudTrail (Splunk Add-on for AWS, sourcetype aws:cloudtrail)``` ```note: Set index= to the index that holds your CloudTrail data.``` ```page: https://threatdeck.net/detections/llmjacking-bedrock-many-keys-one-source-ip/``` ```source: https://sysdig.com/blog/llmjacking-stolen-cloud-credentials-used-in-new-ai-attack/``` ```source: https://permiso.io/blog/exploiting-hosted-models``` ```source: https://docs.aws.amazon.com/awscloudtrail/latest/userguide/query-limitations.html``` index=aws sourcetype=aws:cloudtrail eventSource="bedrock.amazonaws.com" eventType="AwsApiCall" eventName IN ("InvokeModel", "InvokeModelWithResponseStream", "Converse", "ConverseStream") earliest=-7d | stats dc(userIdentity.arn) AS principals dc(recipientAccountId) AS accounts count AS calls values(userAgent) AS user_agents min(_time) AS first_seen max(_time) AS last_seen BY sourceIPAddress | where principals >= 3 | convert ctime(first_seen) ctime(last_seen) | sort - principals - calls ```Claude Code telemetry: Bash tool commands that read credential stores``` ```id: claude-code-otel-bash-touches-secrets | maturity: hunt | techniques: AML.T0055, AML.T0098, ASI02, AML.T0086``` ```MITRE ATT&CK: T1552.001, T1552.004``` ```log source: Claude Code OpenTelemetry log events exported to Splunk (attribute names as documented; OTEL_LOG_TOOL_DETAILS=1)``` ```page: https://threatdeck.net/detections/claude-code-otel-bash-touches-secrets/``` ```source: https://code.claude.com/docs/en/monitoring-usage``` ```source: https://socket.dev/blog/sandworm-mode-npm-worm-ai-toolchain-poisoning``` ```source: https://specterops.io/blog/2025/11/21/an-evening-with-claude-code/``` index=otel event.name IN ("tool_decision", "tool_result") tool_name="Bash" | spath input=tool_parameters path=full_command output=full_command | spath input=tool_parameters path=bash_command output=bash_command | eval cmd=coalesce(full_command, bash_command) | where isnotnull(cmd) | regex cmd="(\.aws/credentials|\.ssh/id_|\.npmrc|\.git-credentials|\.netrc|\.kube/config|\.docker/config\.json|gh\s+auth\s+token|(^|[\s/=])\.env(\s|$|\.))" | stats count AS events values(cmd) AS commands values(decision) AS decisions values(decision_source) AS decision_sources values(permission_mode) AS permission_modes BY user.email session.id | sort - events ```Claude Code telemetry: sessions running in bypassPermissions mode``` ```id: claude-code-otel-bypass-permissions-sessions | maturity: hunt | techniques: LLM06, ASI02, AML.T0053``` ```log source: Claude Code OpenTelemetry log events exported to Splunk``` ```page: https://threatdeck.net/detections/claude-code-otel-bypass-permissions-sessions/``` ```source: https://code.claude.com/docs/en/monitoring-usage``` ```source: https://code.claude.com/docs/en/permission-modes``` index=otel event.name="tool_decision" permission_mode="bypassPermissions" | stats count AS tool_calls dc(tool_name) AS distinct_tools values(tool_name) AS tools min(_time) AS first_seen max(_time) AS last_seen BY user.email session.id terminal.type | convert ctime(first_seen) ctime(last_seen) | sort - tool_calls ```Model output contains markdown images or reference links to external hosts``` ```id: llm-output-markdown-image-exfiltration | maturity: hunt | techniques: AML.T0077, LLM05, LLM02, AML.T0051``` ```MITRE ATT&CK: T1700, T1698``` ```log source: Amazon Bedrock model invocation logs (CloudWatch Logs or S3) in Splunk; text output logging enabled``` ```page: https://threatdeck.net/detections/llm-output-markdown-image-exfiltration/``` ```source: https://simonwillison.net/tags/markdown-exfiltration/``` ```source: https://simonwillison.net/2025/Jun/11/echoleak/``` ```source: https://docs.aws.amazon.com/bedrock/latest/userguide/model-invocation-logging.html``` index=bedrock schemaType="ModelInvocationLog" | spath path=output.outputBodyJson output=model_output | rex field=model_output max_match=20 "!\[[^\]]{0,200}\]\((?:https?:)?//(?[^/)\s\"]+)(?[^)\s\"]*)\)" | rex field=model_output max_match=20 "\[[^\]]{1,100}\]:\s*(?:https?:)?//(?[^/\s\"]+)" | eval dest_host=mvappend(image_host, ref_host) | where isnotnull(dest_host) | eval encoded_param=if(match(mvjoin(image_path, " "), "[?&][^=&\s]{1,40}=[A-Za-z0-9+/=_%-]{32,}"), 1, 0) | mvexpand dest_host | search NOT dest_host IN ("approved-cdn.example.com") | stats count AS outputs dc(requestId) AS requests max(encoded_param) AS encoded_param values(modelId) AS models BY dest_host identity.arn | sort - encoded_param - outputs ```Bedrock guardrail deleted or updated (SPL port)``` ```id: bedrock-guardrail-deleted-or-updated | maturity: hunt | techniques: AML.T0054, LLM01``` ```MITRE ATT&CK: T1685``` ```log source: Splunk: AWS CloudTrail (Splunk Add-on for AWS, sourcetype aws:cloudtrail)``` ```note: Set index= to the index that holds your CloudTrail data.``` ```page: https://threatdeck.net/detections/bedrock-guardrail-deleted-or-updated/``` ```source: https://github.com/SigmaHQ/sigma/blob/master/rules/cloud/aws/cloudtrail/aws_cloudtrail_bedrock_guardrail_deleted.yml``` ```source: https://github.com/SigmaHQ/sigma/blob/master/rules/cloud/aws/cloudtrail/aws_cloudtrail_bedrock_guardrail_updated.yml``` ```source: https://docs.aws.amazon.com/bedrock/latest/userguide/logging-using-cloudtrail.html``` ```Adapted from SigmaHQ rules 59b70e4d-dd17-44a9-b740-acf07ae3eb6a and 1c722651-254a-4b04-a9f4-99b62a2d0a1f (Detection Rule License 1.1)``` index=aws sourcetype=aws:cloudtrail eventSource="bedrock.amazonaws.com" eventName IN ("DeleteGuardrail", "UpdateGuardrail") | table _time eventName requestParameters.guardrailIdentifier userIdentity.arn sourceIPAddress userAgent awsRegion recipientAccountId errorCode ```Burst of Bedrock guardrail interventions from one principal and source (SPL port)``` ```id: bedrock-guardrail-intervention-burst | maturity: hunt | techniques: LLM01, AML.T0054, AML.T0051``` ```MITRE ATT&CK: T1699``` ```log source: Splunk: AWS CloudTrail (Splunk Add-on for AWS, sourcetype aws:cloudtrail)``` ```note: Needs Bedrock guardrail data events (AWS::Bedrock::Guardrail) in the trail that feeds the SIEM; management-event-only trails return nothing. Set index= to the index that holds your CloudTrail data.``` ```page: https://threatdeck.net/detections/bedrock-guardrail-intervention-burst/``` ```source: https://docs.aws.amazon.com/bedrock/latest/userguide/logging-using-cloudtrail.html``` ```source: https://docs.aws.amazon.com/awscloudtrail/latest/userguide/query-limitations.html``` ```source: https://github.com/elastic/detection-rules/blob/main/rules/integrations/aws_bedrock/aws_bedrock_guardrails_multiple_violations_by_single_user.toml``` index=aws sourcetype=aws:cloudtrail eventSource="bedrock.amazonaws.com" eventName="ApplyGuardrail" eventCategory="Data" responseElements.action="GUARDRAIL_INTERVENED" earliest=-7d | bin _time span=1h | stats count AS interventions values(requestParameters.source) AS sources BY _time userIdentity.arn sourceIPAddress | where interventions >= 20 | sort - interventions