# Log ingest targets

Where cloud logs go with each Vulnesis integration, which stay in your account, which are read in place, and the HTTPS endpoints that accept pushed logs.

Two Vulnesis products touch your cloud logs, and they handle them differently:

- **Cloud security** (this segment) reads logs **in place**, inside your account. CloudTrail objects and network logs never leave the account; only findings are reported.
- **Guardian**, the SIEM, takes logs **by push** to an HTTPS endpoint and stores them for search, detection rules and alerts. See [/siem/](/siem.md).

## Read in place by the probe

| Log | Where it lives | Who reads it | Used for |
|---|---|---|---|
| CloudTrail management events | Trail `vulnetix-management-events`, bucket `vulnetix-cloudtrail-<your-account-id>`, in your account | The probe, through `vulnetix-reporter-logdata` | Verification, and seeding the resource ledger from history (30 days by default, at most 90) |
| CloudTrail write events | EventBridge rule `vulnetix-cloudtrail-to-reporter` to queue `vulnetix-security-reporter` | The probe | The resource ledger: which repository or CI session created a resource, and an independent record of resources a sweep failed to enumerate |
| Route 53 Resolver query logs | CloudWatch Logs, in your account | The probe, with the `network-logs` source | Matching names against indicators, last 24 hours |
| VPC flow logs | CloudWatch Logs, in your account | The probe, with the `network-logs` source | Matching addresses against indicators, last 24 hours |

Resolver and flow logs delivered to S3 or Firehose are not read by the probe; the run records the gap `log-destination-not-cloudwatch`. See [Compute runtime](/cloud-security/sources/compute.md#extension-network-logs).

## Pushed to Guardian

All push targets are on `https://ingest.vulnes.is`. Each needs a Guardian ingest token of the matching kind (`aws`, `cloudflare` or `events`), created on the Guardian Sources page and shown once. A token of one kind is refused on another kind's route with `401`.

| Target | Method and path | Auth | Body |
|---|---|---|---|
| VPC flow logs | `POST /v1/ingest/aws/flow` | Firehose access key (`X-Amz-Firehose-Access-Key`) | Firehose HTTP endpoint delivery |
| Route 53 Resolver query logs | `POST /v1/ingest/aws/dns` | Firehose access key | Firehose HTTP endpoint delivery |
| CloudTrail through EventBridge | `POST /v1/ingest/aws/cloudtrail` | Firehose access key | Firehose HTTP endpoint delivery |
| Cloudflare Logpush | `POST /v1/ingest/cloudflare/<dataset>` | `Authorization: Bearer <token>` | Logpush gzip ndjson |
| Your own events, and the runtime sensor | `POST /v1/ingest/events` | `Authorization: Bearer <token>` | `{"source": "...", "events": [{"table": "...", "row": {...}}]}`, at most 5,000 events |

The AWS routes are what the Guardian delivery CloudFormation stack (`vulnetix-siem`) wires up; deploy it per region from the Guardian Sources page. Its CloudTrail stream carries management write events only while a trail logs them, and the cloud security trail `vulnetix-management-events` does. Details: [Guardian AWS log shipping](/siem/sources/aws.md), [Cloudflare](/siem/sources/cloudflare.md), [Events API](/siem/sources/events.md). The runtime sensor forwards to the events route: [Vulnesis sensor](/virtual-patching/runtime/sensor.md).

Guardian needs the Business plan with a Guardian ingest pack. Cloud security needs Pro. The two are independent: you can run either without the other.
