RECON Cloud · CSPM, CIEM, KSPM, SSPM

Cloud security posture,
read-only, in every account

Connect AWS, Azure, Google Cloud, DigitalOcean, Kubernetes, Microsoft 365 and Google Workspace with read-only access. RECON finds the misconfigurations, risky identities and exposed data, shows which ones chain into an attack path, and maps the findings to compliance controls.

  • Read-only, revocable at any time
  • Agentless configuration scan
  • Same 0–10 Risk Score as External and Internal
recon.orizon.one
RECON cloud scan overview for the demo company Meccanica Adda: the AWS connector’s last scan, Risk Score 8.3 out of 10 (High, Grade F), 4 findings, 2 attack paths and 14 resources

Seven connectors, read-only by design

Each connector reads through the provider’s own API with the narrowest access that does the job. RECON never writes to your accounts, and you revoke access from your own console.

  • Amazon Web Services

    Read-only IAM role with External ID

    One CloudFormation stack in your account creates the role. The External ID ties its trust to this connector alone, and RECON assumes the role at scan time, so there are no access keys to store.

  • Microsoft Azure

    Reader + Security Reader

    A Microsoft Entra app with read-only roles on the subscription. The same app reads the Entra ID side: users, roles, PIM eligibility and conditional access. Its client secret is stored encrypted.

  • Google Cloud

    Viewer + Security Reviewer

    A read-only service account created from Cloud Shell, scoped to the project. Its key is stored encrypted.

  • DigitalOcean

    Read-scoped API token

    A personal access token with read scope, plus optional Spaces keys to list object storage. Both are stored encrypted, for this connector only.

  • Kubernetes

    Read-only service account

    The API server URL and the token of a read-only service account, for example one bound to the built-in view role. Nothing is installed in the cluster for the posture scan.

  • Microsoft 365

    One admin consent, read-only Graph

    A Global Administrator approves read-only Microsoft Graph permissions once: directory, policies, app registrations, mailbox settings and SharePoint sharing. No passwords or keys change hands, and access is revoked from the Microsoft admin center.

  • Google Workspace

    Domain-wide delegation, read scopes

    A service account with domain-wide delegation, acting for a super admin you name. It reads users, groups, 2-Step Verification, Gmail forwarding and delegation settings and Drive sharing metadata, never the contents of mail or files.

Every account in one list

Each cloud account, subscription, project, cluster or tenant is one connector. The list shows whether it is online, its latest Risk Score, its resources, findings, critical findings and attack paths, and which data sources it uses.

  • Run a scan or schedule recurring ones from each connector’s menu.
  • Expand a connector for its trends and top findings, without opening the scan.
  • Share a connector with colleagues; only its owner can rename, share or remove it.
  • Remove a connector and the credentials stored for it are destroyed.
recon.orizon.one
RECON cloud connectors page for Meccanica Adda: the AWS, Azure, Google Cloud and Google Workspace connectors, each online, with its Risk Score, resources, findings and attack paths

Five data sources, one always on

The configuration scan needs nothing beyond the read-only access above. Four optional sources go deeper, and each connector shows which ones it uses.

Config
Always on. Agentless reads of every resource, identity and setting through the provider’s API.
Snapshot
A short-lived scanner reads disk snapshots inside your cloud and sends back findings only, never data. On AWS it starts by itself as a Fargate task.
Runtime
An eBPF sensor on your Kubernetes nodes, installed with Helm, streams process, file and network behaviour.
Events
Your cloud event bus (EventBridge, Event Grid or Pub/Sub) forwards control-plane events. A material change re-checks that one resource.
CI/CD
Your pipeline sends infrastructure-as-code changes and container images to be checked before they ship.

Connect, assess, act

  1. Connect

    Choose a provider and grant read-only access in your own console. RECON confirms the grant works before the first scan, which also measures the size of the account.

  2. Assess

    RECON inventories every resource and identity it can read, runs the checks, works out what is reachable from the internet, finds the attack paths and maps the findings to compliance controls.

  3. Act

    Fix in the order RECON sets, findings on a live attack path first. Schedule the next scan, then export the signed HTML report, a PDF, CSV or JSON.

Eleven detection areas, in plain language

Each acronym is a category the security industry uses. Here is what it means, and what RECON actually looks at, grouped by the data each one needs.

From the configuration scan

Agentless. Runs on every scan with nothing more than the read-only connector.

  • CSPMCloud Security Posture Management

    Configuration mistakes in your cloud accounts.

    What RECON checks: Public buckets, snapshots and databases; storage without encryption or TLS; firewalls open to the internet on admin ports; missing audit trails and flow logs; root accounts without MFA; IMDSv1; key and secret rotation; and SPF, DMARC, DNSSEC and CAA on the DNS zones you host.

  • CIEMCloud Infrastructure Entitlement Management

    People and machines that can do too much.

    What RECON checks: Admin-equivalent identities, privilege-escalation paths, stale or long-lived access keys, privileged users without MFA, dormant privileges, cross-account trust, wildcard principals, over-permissive service accounts and Azure PIM roles without MFA.

  • KSPMKubernetes Security Posture Management

    How your clusters and workloads are set up.

    What RECON checks: Public API servers, anonymous access, RBAC switched off, no network policies, outdated versions; privileged or root containers, host namespaces and hostPath mounts, dangerous capabilities, no seccomp, secrets in environment variables, images on :latest and missing Pod Security labels.

  • SSPMSaaS Security Posture Management

    The settings of Microsoft 365 and Google Workspace.

    What RECON checks: MFA and 2-Step Verification for admins, conditional access, legacy authentication, guest and consent settings, risky or unverified OAuth apps, external mail forwarding and delegation, anonymous or external sharing in SharePoint, OneDrive and Drive, and SPF, DKIM and DMARC.

  • AI-SPMAI Security Posture Management

    The AI services and agents in your inventory.

    What RECON checks: Inference endpoints, vector stores, model registries and MCP servers open without authentication; missing guardrails or prompt logging; untrusted or pickle-format models; over-privileged agent tools; sensitive training data; shadow AI. Mapped to the OWASP Top 10 for LLM Applications and the NIST AI RMF.

From the snapshot scan

Looks inside disks. Findings leave your cloud, the data does not.

  • CWPPCloud Workload Protection Platform

    What runs inside your virtual machines.

    What RECON checks: Vulnerable packages ranked with CISA KEV and EPSS, end-of-life operating systems, missing patches, secrets left on disk, malware, and vulnerable hosts that are reachable from the internet.

  • DSPMData Security Posture Management

    Sensitive data kept in a risky place.

    What RECON checks: Personal, card and health data and secrets found in stores that are public, shared across accounts, unencrypted, unlogged, without backup, replicated across borders or kept longer than they should be.

From events and the runtime sensor

Detection of attacks in progress, not only of weak settings.

  • CDRCloud Detection and Response

    Attacks happening in the control plane or on a node.

    What RECON checks: From audit events: logging tampered with, a bucket made public, a firewall opened to the world, admin rights or the owner role granted, a root access key created, console logins without MFA, and the provider’s own threat findings. From the runtime sensor: reverse shells, crypto miners, credential file reads, privilege escalation and container escape.

  • ITDRIdentity Threat Detection and Response

    Signs that an account is being taken over.

    What RECON checks: Impossible travel, MFA fatigue, illicit consent grants and token theft, correlated over a sliding window of each identity’s sign-in, token, consent and audit events.

From your pipeline

The same rules as the live checks, applied before deployment.

  • IaCInfrastructure as Code

    Misconfigurations caught in code, before they go live.

    What RECON checks: Terraform, CloudFormation and Kubernetes manifests, checked with rules that mirror the live AWS, Azure, GCP and Kubernetes checks, plus hardcoded secrets.

  • ImagesContainer image scanning

    What your container images carry.

    What RECON checks: Image CVEs with known-exploited ones first, secrets in layers, images that run as root, unpinned or outdated base images, missing HEALTHCHECK and malicious layers.

Attack paths across identities, network and data

From an exposed resource to the crown jewel

RECON builds a graph of every resource and relationship it reads: what exposes what, who can assume which role, which identity can read or write which store, who belongs to which group. It then walks from each entry point, an internet-exposed resource or an over-privileged third-party grant, to a crown jewel: the account, tenant or project, or a privileged identity.

  • A finding on a live path weighs 1.6 times more than the same finding on its own.
  • On AWS, reachability is checked hop by hop from configuration: public address, route table, network ACL, security group, firewall, load balancer and WAF. A world-open security group in a subnet with no route to the internet is reported as blocked, not exposed.
  • On AWS, effective permissions follow the order AWS itself applies: explicit deny, SCPs, RCPs, permission boundaries, session policies, then grants. Privilege-escalation techniques become edges of the graph.
recon.orizon.one
RECON cloud attack paths for Meccanica Adda: two high paths with risk 7.4, from an internet-facing EC2 instance and from a load balancer to the AWS account

Fix one hop, break the whole path

Each path names the first red hop to fix. Close it and the chain no longer reaches its target, however many findings sit further along. Open a path to see each step it takes to the crown jewel.

  • A path takes the risk of the worst finding on it: 9.2 with a critical, 7.4 with a high, 5.2 with a medium.
  • Paths are listed by severity, and RECON keeps the 25 most serious.
recon.orizon.one
RECON cloud attack path drawer for Meccanica Adda: Path 1 from an internet-facing EC2 instance to the AWS account, with the step that reaches the account and the note that fixing the first red hop breaks the chain

One Risk Score, gated by reachability

Every cloud scan gets a Risk Score from 0 to 10, on the same scale as External and Internal. It weighs each finding by severity and by whether it sits on a live attack path, then applies two rules.

  • A finding on a resource that configuration shows is unreachable from the internet cannot count above medium.
  • A known-exploited (CISA KEV) finding on a reachable admin or crown-jewel resource always counts as critical.
  • The posture score is the same number read the other way: 100 minus ten times the Risk Score.
  • Findings are put in order with a CISA SSVC decision that combines reachability, data sensitivity and exploitation (KEV and EPSS).

Open any finding to see its own risk out of 100 and what makes it: a severity base from 5 to 90, up to 10 more for CVSS and 15 for EPSS, 8 if it faces the internet, 10 on a live attack path, and a floor of 90 when it is known to be exploited (CISA KEV).

recon.orizon.one
RECON cloud misconfigurations for Meccanica Adda: 4 findings, 1 critical, with the finding drawer open on the firewall rule that allows the world on a sensitive port, its risk factors and the compliance controls it fails

Compliance projection

Every finding, mapped to the controls it fails

Each finding carries the benchmark and regulatory controls it fails. RECON rolls them up per framework, so you see which controls are failing and which findings to fix for each. Where a framework publishes a fixed number of controls, the view shows the share with no failing finding.

  • Each failing control opens to show what it checks and the findings behind it.
  • The HTML report counts a control as met only when stored configuration proves a pass, and caps broad organisational frameworks at partially met.
  • Percentages use each benchmark’s published size: 55 controls for CIS AWS, 60 for CIS Microsoft 365, 24 for CIS Kubernetes and 110 for NIST SP 800-53.
  • A projection, not a certification: it shows where configuration falls short of a control, not that an audit is passed.
recon.orizon.one
RECON cloud compliance tab for Meccanica Adda: CIS AWS at 95 percent and NIST 800-53 at 97 percent, three failing controls each, and the table of the six failing controls

20

frameworks and 288 controls in the catalog

MITRE ATT&CK is a technique mapping with no pass rate. HIPAA, GDPR, PCI DSS and FedRAMP have no fixed control count, so they show failing controls rather than a percentage.

Frameworks the checks report against (13)

  • CIS AWS Foundations Benchmark
  • CIS Microsoft Azure Foundations Benchmark
  • CIS Google Cloud Platform Foundation Benchmark
  • CIS Kubernetes Benchmark
  • CIS Microsoft 365 Foundations Benchmark
  • CIS Google Workspace Foundations Benchmark
  • CIS Docker Benchmark
  • NIST SP 800-53 Rev. 5
  • NIST AI RMF 1.0
  • OWASP Top 10 for LLM Applications
  • GDPR
  • PCI DSS v4.0
  • MITRE ATT&CK

Also mapped in the catalog (7)

  • ISO/IEC 27001:2022 Annex A
  • SOC 2 Trust Services Criteria
  • NIS2 Art. 21
  • HIPAA Security Rule
  • CIS Critical Security Controls v8
  • CSA Cloud Controls Matrix v4
  • FedRAMP Moderate

Beside External and Internal, in one RECON

Cloud is a scan type of the same RECON. You start it from the same Create New Mapping window as a domain or an internal scan, and it lands in the same scan list.

  • The same scan header, schedule and export dialog as every other scan type.
  • The same 0–10 Risk Score scale, so a cloud account and a domain are read on one ruler.
  • The same HTML report engine, signed the same way.
  • Connectors can be shared with colleagues; read-only viewers see account IDs and ARNs masked.
recon.orizon.one
RECON’s Create New Mapping window open over the dashboard, with the four scan types: Single Domain, Multiple Domain, Internal and Cloud

A cloud report you can hand over

The overview first, then the plan

The HTML report opens with an exact replica of the scan’s Overview, then explains the findings in plain language, with every number taken from the scan itself.

  • Executive summary, key risks, the attacker’s view, where the risk sits, priorities and a remediation plan in waves: 7, 30 and 90 days, then the next cycle.
  • The projected Risk Score after each wave, labelled as a projection.
  • Compliance status per framework, a map of your regions, the trend against earlier scans, methodology and limitations.
  • In English, Italian or Spanish, with a detached digital signature. Also exported as PDF, CSV (a ZIP) and JSON.
recon.orizon.one
The full technical report of the Meccanica Adda AWS cloud scan, a signed HTML document marked TLP:GREEN: the cover, then the scan’s Overview with Risk Score 8.3

Questions about RECON Cloud

Cloud Security Posture Management is the continuous check of how your cloud accounts are configured: what is public, what is unencrypted, what is not logged and which ports are open to the internet. RECON Cloud does it agentlessly through read-only connectors, and adds identities (CIEM), Kubernetes (KSPM) and Microsoft 365 and Google Workspace (SSPM) in the same scan.
A CNAPP (Cloud-Native Application Protection Platform) is one product that covers cloud posture, identities, workloads, data, Kubernetes, code and threat detection together. RECON Cloud covers those areas, CSPM, CIEM, KSPM, CWPP, DSPM, IaC, container images, CDR and ITDR, and adds SaaS posture for Microsoft 365 and Google Workspace. Unlike a standalone CNAPP, it sits in the same RECON as your external and internal scans.
The configuration scan is: RECON only calls the provider’s API with the read-only access you grant, and nothing is installed. Four optional sources go deeper: a short-lived snapshot scanner that reads disk snapshots inside your cloud and returns only findings, an eBPF sensor for Kubernetes nodes, your cloud event bus and your CI/CD pipeline.
Read-only ones. AWS: an IAM role created by a CloudFormation stack, with an External ID. Azure: an Entra app with Reader and Security Reader. Google Cloud: a service account with Viewer and Security Reviewer. DigitalOcean: a read-scoped token. Kubernetes: the token of a read-only service account. Microsoft 365: one admin consent to read-only Microsoft Graph permissions. Google Workspace: a service account with domain-wide delegation and read scopes.
No. Every connector is read-only and RECON never writes to your accounts. You can revoke access from your own console at any time. Removing a connector in RECON destroys the credentials stored for it and stops any job still in the queue.
Yes. Each AWS account, Azure subscription, Google Cloud project, DigitalOcean team, Kubernetes cluster and Microsoft 365 or Google Workspace tenant is its own connector. They sit together on the connectors page, each with its status, data sources and Risk Score, and each can be scanned on a schedule.
Yes, as SaaS connectors. RECON checks MFA and 2-Step Verification for admins, conditional access, legacy authentication, guest and consent settings, risky OAuth apps, external mail forwarding and delegation, file sharing in SharePoint, OneDrive and Drive, and the SPF, DKIM and DMARC records of your domains.
It gives evidence for the technical measures of NIS2 Article 21, such as access control, MFA, encryption and logging. The compliance catalog includes the ten Article 21(2) measures alongside CIS, NIST and ISO 27001 controls, and the HTML report counts a control as met only when stored configuration proves it. RECON External also tells you whether your company is an essential or an important entity.
There is no list price for Cloud. The price depends on the size of the accounts you connect, which RECON measures on the first scan. Book a demo and we will go through it with you.

See RECON Cloud in a demo

We show you the connectors, the attack paths and the report, and answer your questions on permissions and pricing.