Skip to main content
KibaCloud
Log in

Security

What KibaCloud canand cannot doin your AWS account.

You connect an account by creating a read-only IAM role in it. This page is the full account of what that role permits, what we keep, and how you take it back.

  • Read-only IAM role
  • No stored credentials
  • Metadata only
  • Revoke without asking us

01

We can only read

The role you create carries an enumerated policy written for this product, not the AWS-managed ReadOnlyAccess. Every entry is a List, Describe or Get call. No create, modify or delete action appears anywhere in it, so there is no request KibaCloud could send that would change something in your account.

  • 53 actions, all reads — the complete list is below, not a summary of it
  • Written per collector, so the policy grants what this product actually calls and nothing adjacent
  • The same document is attached to the role in your account

Check it yourself

Open the role in your IAM console and read the policy attached to it.

The policy, in full

All 53 actions the role grants, as they appear in the template you deploy.

  • ec2:Describe*
  • rds:Describe*
  • s3:List*
  • s3:GetBucketLocation
  • elasticloadbalancing:Describe*
  • dynamodb:ListTables
  • dynamodb:DescribeTable
  • dynamodb:ListTagsOfResource
  • elasticache:DescribeCacheClusters
  • elasticache:DescribeReplicationGroups
  • elasticache:ListTagsForResource
  • kafka:ListClusters
  • kafka:DescribeCluster
  • es:ListDomainNames
  • es:DescribeDomain
  • es:DescribeDomains
  • es:ListTags
  • redshift:DescribeClusters
  • ecs:ListClusters
  • ecs:ListServices
  • ecs:DescribeServices
  • ecs:ListTasks
  • ecs:DescribeTasks
  • ecs:DescribeTaskDefinition
  • lambda:ListFunctions
  • lambda:ListTags
  • states:ListStateMachines
  • states:DescribeStateMachine
  • states:ListTagsForResource
  • elasticmapreduce:ListClusters
  • elasticmapreduce:DescribeCluster
  • sagemaker:ListNotebookInstances
  • sagemaker:DescribeNotebookInstance
  • sagemaker:ListEndpoints
  • sagemaker:DescribeEndpoint
  • sagemaker:ListTags
  • rds:ListTagsForResource
  • elasticfilesystem:DescribeFileSystems
  • elasticfilesystem:ListTagsForResource
  • fsx:DescribeFileSystems
  • kinesis:ListStreams
  • kinesis:DescribeStreamSummary
  • kinesis:ListTagsForStream
  • sns:ListTopics
  • sns:ListTagsForResource
  • sqs:ListQueues
  • sqs:GetQueueAttributes
  • sqs:ListQueueTags
  • s3:GetBucketTagging
  • cloudfront:ListDistributions
  • cloudfront:ListTagsForResource
  • apigateway:GET
  • sts:GetCallerIdentity
  • Four entries are wildcards — ec2:Describe*, rds:Describe*, elasticloadbalancing:Describe* and s3:List*. Each expands only over read operations; AWS has no write action behind a Describe or List prefix.
  • apigateway:GET is API Gateway's read action. Its IAM actions are named after HTTP verbs rather than the usual Describe convention, so GET here is a read, and the policy grants no other verb.
  • s3:List* permits listing object keys in a bucket. We list buckets and read their location and tags; no collector enumerates bucket contents.

02

We hold no credentials

The connection record stores a role ARN. We call sts:AssumeRole to obtain short-lived credentials for the duration of a scan, and they expire on their own. There is no long-lived key of yours in our database, which means there is nothing of yours in it to steal.

  • A cross-account role assumed with STS — never an access key, never a secret you paste in
  • A per-account externalId, generated by us and required on every AssumeRole call, which is what closes the confused-deputy problem
  • The role's trust policy names our account and that externalId; nothing else can assume it

Check it yourself

The trust policy sits on the same role. It carries one principal and one condition.

03

Our access cannot grow without you

Because the policy is enumerated rather than a wildcard over services, a new collector needing a permission we do not already hold cannot simply start working. It obliges you to update your stack. There is no path by which we widen our own access in your account.

  • The template declares its own version, and your stack reports that version back when it deploys
  • Each version is published at its own URL and never overwritten, so the URL your stack was created from still serves exactly what you deployed
  • We record the version you deployed, not the one we were offering at the time
  • A connection behind the current version is flagged on your accounts page; an unreported version counts as behind, never as current

Check it yourself

Compare the policy in your IAM console against the template URL your stack was created from.

04

You revoke, and we cannot stop you

Delete the CloudFormation stack in your account and the role goes with it. The next scan fails to assume it and stops. You do not need to tell us first, and there is no setting on our side that keeps access alive after the role is gone.

  • Deleting your workspace stops scanning at that moment — we make no further calls into your account
  • The stack and the role stay in your account until you remove them: we hold no permission to delete either, and the workspace-delete dialog says so rather than implying we cleaned up
  • A role we can no longer assume is reported as access lost, not retried around

Check it yourself

No part of revocation runs through us.

05

What we keep

What we store

  • Resource ARNs and IDs
  • Resource names
  • Tags, exactly as they appear on the resource
  • Region and AWS account ID
  • State — running, available, stopped
  • Per-type metadata: instance class, item counts, desired and running task counts
  • First-seen and last-seen timestamps

What we never store

  • Object contents in S3
  • Rows in your databases
  • Application, access or CloudTrail logs
  • Environment variables, secrets or parameters
  • Security group, VPC and network configuration
  • Anything the resource holds, as opposed to the fact that it exists

The wildcards above permit more read calls than our collectors make — ec2:Describe* covers network configuration we never ask for. Some calls we do make answer with more than we keep: listing Lambda functions returns their environment variables, which we drop without storing. The list on the right is what we never store, not what the policy stops us reading. Where the distinction matters we say which of the two applies.

What the role cannot read at all

  • IAM users, roles and policies
  • CloudTrail
  • Secrets Manager
  • SSM Parameter Store
  • KMS
  • CloudWatch Logs
  • Cost Explorer and billing

Your AWS data is only stored in KibaCloud, nowhere else.

Tags are stored as they are, and we do not inspect them. Teams routinely put a person's name or email in an Owner tag; that is ordinary, and the data processing agreement covers it. It is also a reason not to keep secrets in tag values, on any platform.

06

Where it lives, and who can reach it

One Postgres database, hosted on Neon, encrypted at rest and reached only over TLS. Separation between workspaces rests on two independent controls, and both have to fail before one customer's data reaches another.

  • Postgres row-level security scopes every workspace-owned row to the workspace a request is for
  • A guarded data access layer checks your membership and role before any query runs — row-level security alone scopes by subdomain, not by who is asking, so it is not sufficient on its own
  • Direct database access outside that layer is refused by a lint rule and a test that fails the build, so a new code path cannot skip the check by oversight

Check it yourself

The second control exists precisely because the first is not enough alone.

07

How long we keep it

What is deleted, and when, for each event that ends access
EventWhat happensDetail
You delete a workspaceUnreachable immediately, purged after 30 daysThe window exists so an accidental deletion can still be recovered. Scanning stops at once.
You delete your accountIdentity, credentials, sessions and memberships removed at onceNo retention window. A hashed tombstone of the email survives for one purpose: refusing a repeat free trial.
A subscription endsCollected AWS data purged 30 days after access endsTrial expiry and cancellation alike. The workspace, its members and its audit log are kept until you delete the workspace itself.

What the audit log records

Every workspace keeps its own log. Owners and admins read it in the console.

  • Workspace created
  • Workspace renamed
  • Workspace deleted
  • Subscription activated
  • Plan changed
  • Subscription set to expire
  • Subscription cancelled
  • Member invited
  • Invitation revoked
  • Member role changed
  • Member removed
  • Member left

Data processing

KibaCloud is a processor for the resource data it collects on your behalf. A standard data processing agreement and the current list of sub-processors are available before you sign, in one version for every customer.

Reporting a vulnerability

Mail us. Include what you did and what you saw; a proof of concept helps but is not required. We acknowledge within two business days and tell you what we found, whether or not it turns out to be exploitable. We will not threaten you for reporting in good faith.

security@kibacloud.io