AgentCorruption: AWS Bedrock AgentCore IMDS credential theft
Research disclosure: chat prompts to exposed AgentCore agents could reach IMDS and steal region-wide role credentials, exposing other agents, memories and secrets. AWS has tightened defaults.
What happened
Chat prompt to public AgentCore agent abused tools that could HTTP/shell to IMDS 169.254.169.254, stealing execution-role credentials. Default role was region-wide across AgentCore: invoke other agents, read conversations/memories, pull images, Secrets Manager / AgentCore Identity, plant persistent memory hijacks. AWS tightened IMDSv2 defaults and narrowed default role.
How it happened
Read the flow as text & sources
A catalog reconstruction of the supplied source account. Unknown steps remain explicit; an edge describes the reported relationship between steps.
- 01 · Public agent receives a prompt
An exposed AWS AgentCore agent accepts attacker input.
Hunt: An agent ingests external content or an untrusted session identifier before a tool action or configuration change outside the approved task.
Requires application instrumentation and a custom field mapping. Input exposure does not prove that the model followed instructions, executed a tool or bypassed approval. Avoid collecting sensitive prompt bodies unless authorized.
Source for step 1 - 02 · Agent tools reach metadata
Prompt injection abuses HTTP or shell tools to query the instance metadata service.
Hunt: An agent tool reaches a metadata endpoint and the resulting role identity is used outside its intended task or resource scope.
Metadata traffic may stay on the host and be absent from ordinary proxy logs. IMDS access can be legitimate; correlate the role and task.
Source for step 2 - 03 · Execution credentials are exposed
The response discloses the agent execution-role credentials.
Hunt: A service key or role is used from an unfamiliar origin, gains broader access or causes inference usage outside its established baseline.
Cloud control-plane logs alone may not include model prompts or inference calls. No endpoint policy is claimed to stop cloud-only key misuse.
Source for step 3 - 04 · Broader agent access
The role can reach other agents, memories and connected services; persistent memory hijacks are described.
Hunt: A service key or role is used from an unfamiliar origin, gains broader access or causes inference usage outside its established baseline.
Cloud control-plane logs alone may not include model prompts or inference calls. No endpoint policy is claimed to stop cloud-only key misuse.
Source for step 4
Why it matters
Reported entry point: Prompt injection / SSRF via agent tools against exposed AgentCore agent. Reported impact: The role can reach other agents, memories and connected services; persistent memory hijacks are described; reported targets include Organizations deploying AgentCore agents (research disclosure; patched).
Check these source-linked conditions against your setup.
The research concerns AWS Bedrock AgentCore agents with tools that can reach instance metadata.
Conditions to check- Confirm use of AgentCore, exposure to untrusted prompts and HTTP / shell tool access to metadata.
- Review the execution-role permissions and the defaults / mitigations described in the source.
Source-linked applicability guidance. Confirm current scope in the original report.
What you can do
Protection & prevention
Review AgentCore metadata access and least-privilege execution roles; check the current vendor guidance.
Least-privilege agent IAM; IMDSv2; agent's role is its blast radius
Catalog defensive guidance based on the reported behavior. Confirm scope and current mitigations in the original source.
Detection queries
Draft query starters grouped by the attack stages where they apply.
Start with the logs you collect. Each hunt identifies its required visibility, field mapping and limits. Queries are catalog-authored drafts using reusable behavior profiles; test with known activity and your normal baseline before alerting.
01Public agent receives a promptUntrusted content enters an agent sessionCustom schema · Draft
An exposed AWS AgentCore agent accepts attacker input.
What to look for
An agent ingests external content or an untrusted session identifier before a tool action or configuration change outside the approved task.
Collect the exposed agent input and request/session context. A received prompt alone does not establish exploitation.
Visibility: Application / agent audit records
Required logs
- Input provenance, source type and request/session identifier
- Captured tool responses or retrieved content where collection is approved
- Subsequent tool calls, configuration changes and approval decisions
Coverage limit. Requires application instrumentation and a custom field mapping. Input exposure does not prove that the model followed instructions, executed a tool or bypassed approval. Avoid collecting sensitive prompt bodies unless authorized.
Custom-schema candidate for input exposure only. Map agent.input.ingest and agent.session.resume to your application request and session logs. CorrelationId must link subsequent actions. EventType values are proposed normalization names, not native vendor events.
Sigma
title: Hunt - Untrusted content enters an agent session
id: 2d89d7ae-9d44-4dc6-a95b-f9c8de7b8352
status: experimental
description: An agent ingests external content or an untrusted session
identifier before a tool action or configuration change outside the approved
task.
author: MagicSword OpenAttacks catalog
date: 2026-11-04
falsepositives:
- Authorized activity matching this broad hunt; baseline and investigate
before alerting.
level: low
logsource:
category: application
product: openattacks_audit
definition: Custom normalized audit schema; map EventType, Actor, Asset, Action,
Target and CorrelationId from the required source logs first.
detection:
selection:
EventType:
- agent.input.ingest
- agent.session.resume
condition: selection
KQL
// SCHEMA TEMPLATE: create/map this custom table before use; not a built-in Sentinel table.
// OpenAttacksAudit_CL: TimeGenerated(datetime), EventType/Actor/Asset/Action/Target/CorrelationId(string).
// Selects candidates only. Apply the stage hypothesis and correlation steps before alerting.
OpenAttacksAudit_CL
| where TimeGenerated > ago(7d)
| where EventType in ("agent.input.ingest", "agent.session.resume")
| project TimeGenerated, EventType, Actor, Asset, Action, Target, CorrelationId
| order by TimeGenerated descSPL
sourcetype="openattacks:audit"
EventType IN ("agent.input.ingest", "agent.session.resume")
| table _time EventType Actor Asset Action Target CorrelationId
| sort - _timeInvestigation steps
- Join the ingested input to later actions by session or request ID; preserve the original content as evidence.
- Compare the resulting action with the approved task and baseline. Receiving suspicious text alone is not successful injection.
02Agent tools reach metadataAgent tool access to cloud metadataCustom schema · Draft
Prompt injection abuses HTTP or shell tools to query the instance metadata service.
What to look for
An agent tool reaches a metadata endpoint and the resulting role identity is used outside its intended task or resource scope.
Visibility: Defender-side logs where collected
Required logs
- Agent HTTP / shell tool audit
- Workload egress / metadata access instrumentation
- Role-session and downstream cloud audit
Coverage limit. Metadata traffic may stay on the host and be absent from ordinary proxy logs. IMDS access can be legitimate; correlate the role and task.
Custom-schema candidates, not out-of-box detectors. Map your application / cloud / platform logs to EventType, Actor, Asset, Action, Target and CorrelationId. KQL uses a custom OpenAttacksAudit_CL table; SPL uses openattacks:audit. EventType values are proposed normalization names, not native vendor event names. Correlation and environment-specific alert logic remain to be built.
Sigma
title: Hunt - Agent tool access to cloud metadata
id: c88eca38-d2f7-4e6a-aabf-6a58f5787823
status: experimental
description: An agent tool reaches a metadata endpoint and the resulting role
identity is used outside its intended task or resource scope.
author: MagicSword OpenAttacks catalog
date: 2026-10-23
falsepositives:
- Authorized activity matching this broad hunt; baseline and investigate
before alerting.
level: low
logsource:
category: application
product: openattacks_audit
definition: Custom normalized audit schema; map EventType, Actor, Asset, Action,
Target and CorrelationId from the required source logs first.
detection:
selection:
EventType:
- agent.metadata.request
condition: selection
KQL
// SCHEMA TEMPLATE: create/map this custom table before use; not a built-in Sentinel table.
// OpenAttacksAudit_CL: TimeGenerated(datetime), EventType/Actor/Asset/Action/Target/CorrelationId(string).
// Selects candidates only. Apply the stage hypothesis and correlation steps before alerting.
OpenAttacksAudit_CL
| where TimeGenerated > ago(7d)
| where EventType in ("agent.metadata.request")
| project TimeGenerated, EventType, Actor, Asset, Action, Target, CorrelationId
| order by TimeGenerated descSPL
sourcetype="openattacks:audit"
EventType IN ("agent.metadata.request")
| table _time EventType Actor Asset Action Target CorrelationId
| sort - _timeInvestigation steps
- Look for metadata destination access from the agent's execution context and determine which credentials were available.
- Trace that role session into agent memory, secrets and other resources; review the permissions boundary.
03Execution credentials are exposedUnexpected cloud / inference identity useCustom schema · Draft
The response discloses the agent execution-role credentials.
What to look for
A service key or role is used from an unfamiliar origin, gains broader access or causes inference usage outside its established baseline.
Visibility: Defender-side logs where collected
Required logs
- Cloud control-plane and identity audit
- AI-service invocation / diagnostic logs when enabled
- Usage, quotas, cost and key inventory
Coverage limit. Cloud control-plane logs alone may not include model prompts or inference calls. No endpoint policy is claimed to stop cloud-only key misuse.
Custom-schema candidates, not out-of-box detectors. Map your application / cloud / platform logs to EventType, Actor, Asset, Action, Target and CorrelationId. KQL uses a custom OpenAttacksAudit_CL table; SPL uses openattacks:audit. EventType values are proposed normalization names, not native vendor event names. Correlation and environment-specific alert logic remain to be built.
Sigma
title: Hunt - Unexpected cloud / inference identity use
id: afac2e65-c838-485b-a26f-7f175d422f61
status: experimental
description: A service key or role is used from an unfamiliar origin, gains
broader access or causes inference usage outside its established baseline.
author: MagicSword OpenAttacks catalog
date: 2026-10-23
falsepositives:
- Authorized activity matching this broad hunt; baseline and investigate
before alerting.
level: low
logsource:
category: application
product: openattacks_audit
definition: Custom normalized audit schema; map EventType, Actor, Asset, Action,
Target and CorrelationId from the required source logs first.
detection:
selection:
EventType:
- cloud.credential.use
- cloud.role.change
- model.invoke
condition: selection
KQL
// SCHEMA TEMPLATE: create/map this custom table before use; not a built-in Sentinel table.
// OpenAttacksAudit_CL: TimeGenerated(datetime), EventType/Actor/Asset/Action/Target/CorrelationId(string).
// Selects candidates only. Apply the stage hypothesis and correlation steps before alerting.
OpenAttacksAudit_CL
| where TimeGenerated > ago(7d)
| where EventType in ("cloud.credential.use", "cloud.role.change", "model.invoke")
| project TimeGenerated, EventType, Actor, Asset, Action, Target, CorrelationId
| order by TimeGenerated descSPL
sourcetype="openattacks:audit"
EventType IN ("cloud.credential.use", "cloud.role.change", "model.invoke")
| table _time EventType Actor Asset Action Target CorrelationId
| sort - _timeInvestigation steps
- Pivot on key ID / principal, source network and user agent, then inspect role changes, token minting and service pivots.
- Join inference usage and billing anomalies to the same principal; verify whether the behavior was authorized.
04Broader agent accessUnexpected cloud / inference identity useCustom schema · Draft
The role can reach other agents, memories and connected services; persistent memory hijacks are described.
What to look for
A service key or role is used from an unfamiliar origin, gains broader access or causes inference usage outside its established baseline.
Visibility: Defender-side logs where collected
Required logs
- Cloud control-plane and identity audit
- AI-service invocation / diagnostic logs when enabled
- Usage, quotas, cost and key inventory
Coverage limit. Cloud control-plane logs alone may not include model prompts or inference calls. No endpoint policy is claimed to stop cloud-only key misuse.
Custom-schema candidates, not out-of-box detectors. Map your application / cloud / platform logs to EventType, Actor, Asset, Action, Target and CorrelationId. KQL uses a custom OpenAttacksAudit_CL table; SPL uses openattacks:audit. EventType values are proposed normalization names, not native vendor event names. Correlation and environment-specific alert logic remain to be built.
Sigma
title: Hunt - Unexpected cloud / inference identity use
id: afac2e65-c838-485b-a26f-7f175d422f61
status: experimental
description: A service key or role is used from an unfamiliar origin, gains
broader access or causes inference usage outside its established baseline.
author: MagicSword OpenAttacks catalog
date: 2026-10-23
falsepositives:
- Authorized activity matching this broad hunt; baseline and investigate
before alerting.
level: low
logsource:
category: application
product: openattacks_audit
definition: Custom normalized audit schema; map EventType, Actor, Asset, Action,
Target and CorrelationId from the required source logs first.
detection:
selection:
EventType:
- cloud.credential.use
- cloud.role.change
- model.invoke
condition: selection
KQL
// SCHEMA TEMPLATE: create/map this custom table before use; not a built-in Sentinel table.
// OpenAttacksAudit_CL: TimeGenerated(datetime), EventType/Actor/Asset/Action/Target/CorrelationId(string).
// Selects candidates only. Apply the stage hypothesis and correlation steps before alerting.
OpenAttacksAudit_CL
| where TimeGenerated > ago(7d)
| where EventType in ("cloud.credential.use", "cloud.role.change", "model.invoke")
| project TimeGenerated, EventType, Actor, Asset, Action, Target, CorrelationId
| order by TimeGenerated descSPL
sourcetype="openattacks:audit"
EventType IN ("cloud.credential.use", "cloud.role.change", "model.invoke")
| table _time EventType Actor Asset Action Target CorrelationId
| sort - _timeInvestigation steps
- Pivot on key ID / principal, source network and user agent, then inspect role changes, token minting and service pivots.
- Join inference usage and billing anomalies to the same principal; verify whether the behavior was authorized.
Technical detailsTooling, impact & MITRE ATLAS
Attack context
- AI's role
- Agent / MCP as target
- Initial access
- Prompt injection / SSRF via agent tools against exposed AgentCore agent
Tooling as reported
AI products, models and artifacts
Other tooling
- IMDS / IMDSv2
- AWS Secrets Manager
- ECR
Targets and reported impact
- Targets / sectors
- Organizations deploying AgentCore agents (research disclosure; patched)
- Reported scale / exposure
- Single prompt to one agent → account/region AgentCore blast radius (pre-fix)
Reported attribution
Vulnerability research (Zenity at SecTor)
Structured attribution
- Actor
- Vulnerability research (Zenity at SecTor)
Detection rules
- Review attack-specific indicatorsGUIDANCE
Monitor for the techniques described in AgentCorruption (AWS Bedrock AgentCore IMDS). Refer to sources for specific IOCs and TTPs.
Source defensive notes
Least-privilege agent IAM; IMDSv2; agent's role is its blast radius
MITRE ATLAS
AI-specific catalog annotations based on the reported behavior; these are not MITRE case-study assignments.
A prompt sent directly to the public agent redirects its tool use. Source
The metadata response exposes execution-role credentials. Source
The source describes persistent memory hijacks. Source
Sources & research credit
1 sourceResearch by Tamir Ishay Sharbat, Lana Salameh, Dmitry Lozovoy · Zenity Labs
Record & downloads
- Source record name
- AgentCorruption (AWS Bedrock AgentCore IMDS)
- Entry ID
- agentcorruption-bedrock-2026-10
- Catalog review
- October 20, 2026
- Category
- attacks-on-ai
- AI involvement confidence
- confirmed