Back to evidence

Sanitized live incident

Confirmed compromise

Native source identity and targetable endpoints are private.

criticalconfirmed
Confidence
100%
First seen
Aug 30, 8:44:11 AM PDT
Evidence through
Aug 30, 8:49:24 AM PDT
AI status
Complete
True positive100% confidence

True positive confirmed compromise: captured exploit/response evidence shows command-oriented input followed by non-reflected root identity and kernel output in the responding workload, despite HTTP 400 [[redacted]; [redacted]]. Event-driven telemetry independently observed root shells and discovery processes in the correlated workload [redacted]. The detector's immutable state is confirmed, and the inspected evidence supports that result. Later sensitive-target and network-client-class processes increase concern, but request-to-process and request-to-socket causality remain unproven [redacted].

Attack stage
Successful exploitation/root execution with correlated shell and discovery activity
Model
gpt-5.6-sol · 13 evidence calls

Observed impact

  • Remote command execution as root in the responding workload, with process identity and system information exposed [[redacted]; [redacted]].
  • Root shell and discovery processes executed in the correlated workload [redacted].
  • A root shell classified as targeting a sensitive resource was observed; actual file content access or disclosure is not established [redacted].
  • A root shell classified as network-client capable and a root bash process were observed, but no outbound connection is established [redacted].
  • A PATCH request received HTTP 200 after compromise, but authorization and an actual state mutation are not established [[redacted]].

Deterministic signals

Http.server command output100%

Exploit request received non-reflected process identity output

2 observations · 2 http
Process.correlated discovery command80%

A discovery process appeared in the correlated workload and request window

6 observations · 6 process
Process.correlated shell spawn80%

A shell process appeared in the correlated workload and request window

17 observations · 12 process
Process.correlated exit99%

A previously correlated process lifecycle exited

19 observations · 12 process
Http.command injection attempt88%

Request contains shell metacharacters and command tokens

8 observations · 8 http
Http.server kernel output97%

Response contains non-reflected kernel identification

1 observations · 1 http
Process.correlated network client80%

An outbound-capable client process appeared in the correlated workload and request window

1 observations · 1 process
Process.correlated sensitive file command80%

A process command targeted a sensitive file in the correlated workload and request window

1 observations · 1 process
Http.state changing activity after compromise88%

A state-changing HTTP method was observed after confirmed workload compromise

1 observations · 1 http

Explicit uncertainty

  • Process correlation is based on workload and temporal proximity; there is no unique request-to-process causality edge.
  • The incident cites no flow-plane evidence. The network-client process therefore cannot establish a connection, destination, command-and-control, or exfiltration.
  • The sensitive-target classification does not show whether file contents were successfully read, modified, or returned.
  • The PATCH response does not establish caller authorization, credential theft, or a durable state change.
  • No cited evidence proves host escape, persistence, lateral movement, or data theft; proven execution is scoped to the responding workload.
  • The source key is a traffic/workload cluster and may represent a proxy, NAT gateway, or multiple workers rather than one actor.
  • Exact process exits establish completion of those lifecycles only; they do not show that the workload is clean or that access ended.

Recommended actions

  1. Immediately isolate or remove the affected workload from service while preserving its runtime, application, and orchestrator evidence; redeploy from a known-good image after investigation.
  2. Identify and remediate the vulnerable API input path, including strict input validation, safe command invocation, and regression tests for shell metacharacter injection.
  3. Review workload and orchestrator audit telemetry across the incident window for filesystem changes, new processes, secret access, persistence, container escape, and activity in adjacent workloads.
  4. Review egress, DNS, proxy, and network-flow logs for the affected workload around the network-client process time; do not infer outbound communication from process class alone.
  5. Determine which sensitive resource was targeted, verify its integrity and access history, and rotate any credentials or secrets shown to have been exposed.
  6. Audit the post-compromise PATCH transaction and associated authentication/session history to determine whether it was authorized and whether state actually changed.
  7. Reduce blast radius by running the workload as a non-root identity with least privilege, a read-only filesystem where feasible, and constrained egress.
  8. Treat source-cluster blocking as a temporary risk-control decision rather than identity attribution; investigate upstream proxy or NAT context before assigning an actor.