One of the most reasonable questions a security-conscious buyer asks: "Your scanner is shared across customers. What stops tenant A from seeing tenant B's data?"
The short answer: logical isolation — per-tenant credentials and tenant-scoped data. There is no dedicated, single-customer scanner option today. This post walks through each risk a shared scanner introduces and what we do about it, including the parts that are plain rather than impressive.
How scanning actually runs
Cloud scans run as scheduled jobs on three cadences: a 5-minute pass (15 minutes during a trial) that reads findings the cloud provider has already raised (Defender, GuardDuty, Security Hub, the Azure Activity Log, service health), an hourly pass over the exposure and state scanners, and a nightly pass over everything, including expiry, rotation and cost. Each job starts, works through the active connections (Azure one after another; AWS and Google Cloud up to eight at a time), writes its findings and exits; nothing sits running in between. Endpoint and certificate checks are a separate job that runs every 5 minutes; a paid plan is checked on every run, a trial every 15 minutes. A customer can ask for a fresh scan of a connection, but only once every 15 minutes per account, and it waits for the next hourly pass, so one tenant cannot crowd out another by asking for more. While CloudMonitor is pre-launch, the service is powered down for part of each day and nothing scans while it is down; the Terms say so too.
The real risks — and what stops each one
1. Credential leakage
The risk: the scanner holds tenant A's cloud credentials while scanning. A bug or sloppy logging could expose them.
What we do: where the provider allows it, we avoid holding a secret at all — the secretless Azure connection (Beta, workload identity federation) needs none (see our auth-method comparison). Today AWS connections use a read-only IAM access key and GCP connections a service-account key; cross-account roles and workload identity federation are where we want to get to, and are not built yet. Stored credentials are encrypted at rest with AES-256-GCM under one application-level key held outside the database (not a key per tenant), decrypted only for the scan that needs them, and never written to logs.
2. Scan output written to the wrong tenant
The risk: the scanner correctly runs tenant A's scan but the result lands under tenant B.
What we do: a finding is written against the connection it came from, and the connection belongs to exactly one tenant. Every dashboard query, export and API response is scoped to the signed-in tenant.
3. Noisy neighbor
The risk: tenant A connects 2,000 resources and their scan delays everyone else's.
What we do: honestly, not much yet, because it has not been needed: connections are worked through inside each scheduled window (Azure one after another, AWS and Google Cloud up to eight at a time, shared by every tenant), so a very large tenant makes the pass longer rather than starving anyone. There are no per-tenant queues or concurrency caps. If the nightly pass ever stops fitting in the night, that is the point at which we build them.
4. Data-plane vs. metadata-plane
The risk: even with credentials isolated, a scanner with broad permissions could read sensitive contents — secret values, database rows, log bodies.
What we do: the credentials we ask you to grant are metadata-plane only. Reader / Monitoring Reader / Viewer roles let us see that a Key Vault has 12 secrets and one of them expires next week; they do not let us read the value of any secret. Each connector post lists what its scanners read.
5. Our own access
The risk: the vendor's staff.
What we do: say it plainly. The operator's staff roles can read tenant data in order to run the service, diagnose faults and answer support requests. Section 9 of the Privacy Policy sets out what that covers.
What we cannot offer
- No SOC 2, ISO 27001 or other certification or attestation.
- No SLA.
- No dedicated scanner, no per-tenant encryption keys, no bring-your-own-key.
If your regulator or your own policy requires any of those, CloudMonitor is not the right fit today, and we would rather you heard that from us. If you have a concern this post didn't cover, tell us.