CIEM — AWS Risk Dashboard & IAM
Turning a wall of disconnected AWS findings into one place to see every identity risk on the account — and fix it, without leaving the product.
Overview
This CIEM module scans an AWS account for identity and access risk — every IAM user, group and role, every EC2 instance role, every S3 bucket and RDS instance — and turns what the scan finds into something a security or platform team can actually act on.
Before this module, findings lived wherever the scan put them — no shared severity model, no way to tell which risks sat on identities that mattered, and no path from "here's a risk" to "here's it fixed" without leaving the product.
One dashboard that shows the account's total exposure at a glance, lets anyone drill from a single number down to the exact identity behind it, and — since the platform already owns the policy layer — lets that person remediate the finding on the spot.
Who this is for
Two roles open this dashboard, at a different pace and for a different reason. The screens are designed around that gap.
Cloud Security Engineer
Primary · dailyOwns the AWS account's security posture day to day. Opens the dashboard right after a new scan lands, or when an alert points here.
Triage new Critical findings first · know what changed since the last scan · fix what's fixable without leaving the tool.
Findings with no severity or owner · identity context scattered across AWS console tabs · a "fix" that means filing a ticket and waiting.
Security / Compliance Manager
Secondary · periodicAnswers for the account's posture upward — to leadership, to an auditor. Opens weekly, or before a review, and rarely drills into a single identity.
A defensible snapshot of exposure · proof a flagged risk actually got fixed · a trend, not just a point-in-time count.
Totals that don't reconcile with categories · no record of what changed between scans · severity that reads as a color with no definition.
Shaped in working sessions with the product owners and security engineers who use this daily — the "vs. previous scan" trend and the Unscored bucket later in this case exist directly because of the Manager persona's frustrations above.
Competitive landscape
CIEM is still a young, consolidating category. Three products get closest to the same problem: turning entitlement sprawl into something a person can act on.
| Product | How it shows risk | Remediation path | Note |
|---|---|---|---|
| Microsoft Entra Permissions Management (retired 2025) | Permission Creep Index — Low/Medium/High across AWS, Azure, GCP. | In-product dashboard — delete or modify permissions and roles directly, approve on-demand access requests. | Closest naming match to "permission creep" — pulled from sale in 2025. |
| Wiz — CIEM Explorer | Security Graph mapping every identity to reachable resources. | Configurable — one-click to fully automated. Reversible. | Furthest on automation, across a full spectrum from manual approval to fully autonomous. |
| CyberArk — Cloud Entitlements Manager | Least-privilege recommendations across AWS, Azure, GCP. | Terraform / IaC pipeline — applied by a developer via terraform apply. | Built by a PAM vendor, with remediation routed through infrastructure code rather than the product itself. |
Wiz and CyberArk have offered this kind of finding-to-fix visibility for a while now; Microsoft only pulled its own version from sale in 2025. That's the practical reason this module exists in the platform — not to leapfrog the category, but so the product wasn't visibly behind on a capability customers already expect from a modern PAM suite.
Understanding the risk model
Before any screen, the risk model needed a shape everyone could agree on — working with the security engineers who built the scanning logic.
A fourth state for findings the policy hasn't scored yet — kept visible everywhere, not folded into the total, so counts always reconcile.
Every finding rolls up into one category, each with its own specific finding types.
- Public Exposure
- Identity & Entitlements
- Inactive Identities
- Network Exposure
- Weak Auth & Credentials
- Unencrypted Resources
IAM users and groups on one side; IAM roles and EC2 instance roles on the other. This split mattered more than any single category — machine identities carry most of the risk and are the ones nobody checks by habit.
One loop, two surfaces
Designed so a person never has to leave the product, or even the current screen, to go from a number to a fix.
See what the account looks like now
Total risk, top findings, resource counts and the six categories — read-only, compared against the previous scan.
Turn a risk type into a filter
Risk-type cards sit above the identity table — selecting one narrows the table live. Count and rows are never two places.
Go from identity to evidence
Used-vs-granted access, services and policies, and every risk on that identity — in a single read-only drawer.
Fix it without switching tools
Runs through the platform's own PAM policy sync — no AWS console, no ticket; revoked access stays requestable through JIT.
Key screens
The loop above, screen by screen — the real interface end to end, from the dashboard's total risk down to the confirmation that closes a finding.
Dashboard — total risk and top findings
The account's exposure in one ring — Critical, High, Low and Unscored — next to findings ranked by severity, each tagged with the resource type and count behind it.
Entitlement posture & human vs. non-human
Granted-vs-used access broken down by identity type, next to the split that mattered most — 52% of non-human identities carried a finding, against 7% of human ones.
Risk by category
All findings grouped into six categories, each with its own severity mix and top finding types — the same structure the IAM, EC2, S3 and RDS pages inherit.
From a top finding to the affected roles
Selecting a finding on the dashboard opens a read-only, exportable list of every entity behind it — here, the IAM roles carrying at least one risk.
IAM — risk cards filter the table beneath them
Fourteen risk types as cards above the identity table; selecting one (or several) narrows the rows live, with severity always shown as a labeled chip, not color alone.
Identity detail — used vs. granted access
The case for revoking access made in one sentence before any table — 18 of 24 granted services on this identity have never been used.
Remediation, confirmed in plain language
Scope and action spelled out, a required checkbox, and the line that matters most: revoked entitlements stay requestable through PAM's JIT access.
From a risk card to bulk remediation
"View details" on any card opens the affected entities behind that risk — select as many as needed and remediate or share the finding in one action.
Every risk on a row, without leaving the table
The "+1 more" chip expands in place into one card per risk — reason, severity and all — instead of a jump to a separate drawer.
A dead end that explains itself
A no-match state states plainly what happened, why, and the one action back — the three things empty-state guidance flags as most often skipped.
Design decisions
Severity never rests on color alone
Every chip carries its label — Critical, High, Low — independent of color, across every table and drawer. Replaced an earlier icon-only version once tables ran into hundreds of rows.
Unscored findings get their own bucket
Rather than disappear from the total, findings without an assigned severity show as their own share of it — so the ring always reconciles with the count above it.
Human vs. non-human is a comparison, not a filter
Framed as its own panel because non-human identities carried the majority of the risk — a gap that would stay invisible inside a single identity table. Added in a later pass.
Read-only until that one explicit action
Export, copy resource IDs, and a persistent "read-only view" label across every drawer — browsing thousands of findings should never risk changing one by accident.
Cards double as filters, not just stats
Selecting a risk-type card narrows the table beneath it live — the number and the list are never two separate places to check. Started single-select; became multi-select in a later pass.
Two ways into the same detail, on purpose
A risk card's "View details" opens every entity behind one risk type; a row's "+1 more" opens every risk on one identity. Same data, read from whichever direction the question started.
The case for action comes before the table
"18 of 24 granted services never used" reads as a sentence and a bar before a single row of policy data — the decision to remediate shouldn't require reading a table first.
Remediation is a confirmation, never a silent click
Scope, action and a required acknowledgment — revoking live AWS permissions is not something anyone should be able to do by accident. Added last, once the flow felt safe enough to ship.
Reflection
This was designed solo, working directly with the product owners and security engineers who defined the risk logic behind it — the harder part was less the interface than agreeing, screen by screen, on what a "risk" was allowed to mean.
These screens just shipped, so two things stay open: extending this same pattern to EC2, S3 and RDS, and — the one that actually matters — seeing how security teams use it once real feedback starts coming in.
What's on this page differs from what actually shipped, too — the screens were reworked to mask the real product, and a few pieces went further while I was already in there: multi-select filtering, the previous-scan comparison, and remediation itself go beyond what time and technical constraints let the live version include. They're additions, not proof the pattern's been validated in production yet — that still depends on the feedback in the point above.