Enforced against intended
What an authorisation review compares, line by line
A review is a comparison between two things: the access the systems actually grant, and the policy they were meant to enforce. The gaps between them are the findings, and each row here is a form that gap takes.
The roles and what each grants
The job each role is meant to serve and the least access that job needs.
The gap it reveals: Role explosion — dozens of overlapping roles nobody can reason about, granting far more than the job requires.
The accounts and their live grants
The people and services that are meant to still exist and still need that access today.
The gap it reveals: Orphaned permissions — access left behind by a leaver, a finished project or a retired integration.
The access one ordinary user has to another's data
The rule that a user may reach only their own records, enforced on the server for every object.
The gap it reveals: A horizontal escalation path — a way to reach a peer's data by asking for it directly.
The route from an ordinary role to a privileged one
The rule that elevating privilege requires an authorised, audited action, not a side effect.
The gap it reveals: A vertical escalation path — a way for a low-privilege account to gain a high-privilege one.
How the model rots
How role explosion and orphaned permissions accumulate unnoticed
No one designs a broken access model; it accretes. A new team needs slightly different access, so rather than reason about the existing roles someone clones one and adjusts it, and now there are two nearly identical roles. Repeat that over years and there are forty, overlapping in ways no document captures, each granting a little more than its holders need because it was easier to copy a generous role than to define a precise one. This is role explosion, and its danger is not any single grant but that the model becomes impossible to reason about, so nobody can say with confidence who can do what.
Orphaned permissions accumulate the same way, from the opposite direction: access is granted readily and revoked reluctantly. A person changes team but keeps the old access alongside the new; a contractor leaves but the service account they used stays live; a project ends but the integration's standing permissions remain. Each is individually explicable and collectively a growing surface of access that maps to nobody's current need. A review catches these not by testing each in isolation but by comparing the whole picture — every grant that exists — against the whole policy of who is meant to have what, which is the only way the ones nobody remembers surface.
The escalation path
How a small foothold becomes control when the model has gaps
The most serious authorisation findings are not single over-grants but chains: a sequence of small, individually minor gaps that together carry an account from where it started to somewhere it should never reach. The drawing follows one such chain.
01
A low-privilege account
An ordinary user, or a compromised one, with only the access its everyday job needs — the intended starting point.
- User role
02
A horizontal step
A missing object-level check lets it read data belonging to other users at the same level, widening what it can see without yet raising its rank.
- Peer records
03
A misplaced grant
Among the newly visible data is something that should have been privileged — a configuration value, a token, a role assignment left editable by an over-broad permission.
- Orphaned grant
04
A vertical step to control
Using that, the account assigns itself a higher role or acts as one, completing the climb from ordinary user to administrator that no single gap would have allowed alone.
- Admin role
Fix the model, not the step
Patching only the final step leaves the chain one gap short of working again. The review reports the whole path and the model change — a re-checked object boundary, a permission scoped down, a role assignment made privileged — that removes the class, not the instance.
A single over-grant is a bug; a path from user to administrator is a model that needs correcting.
Our role
What the review measures, and why the policy must come from you
- 01
We map the access as it really is
We read every role, account and grant across the estate into one picture, because you cannot correct a model you cannot see, and the live grants almost never match the diagram anyone drew.
- 02
We measure against your policy
A finding is a gap between enforced and intended access, so the intended side — who is meant to have what — has to come from you. Where it is not written down, we help you make it explicit, but we measure against your intent rather than substitute our own.
- 03
We report the model, not just the bugs
Each escalation path comes with the model-level change that closes its class, so you fix the structure that produced the finding rather than the one instance we happened to trace, and your team owns the change to how access is granted from then on.
The intended policy — who should be able to do what — is yours to set, and the priority order for closing the gaps is your risk decision; we map the access as enforced, find where it diverges from your policy, and report the model changes that keep it from drifting back.
Common questions
What buyers ask
How is an authorisation review different from the access certification our IAM tool already runs?+
An access certification asks managers to confirm that the grants their people hold look right, which is useful but circular: it checks the current state against people's memory of why it exists, not against a policy of what it should be, and a manager rubber-stamping a long list rarely spots that a permission is one link in an escalation chain. An authorisation review measures the enforced access against an intended policy and, crucially, tests whether the systems actually stop what the policy forbids — whether an ordinary account can in fact reach a peer's data or climb to a privileged role. Certification confirms grants are approved; the review confirms the enforcement holds and the model is sound. They answer different questions, and the escalation paths live in the gap the certification does not look at.
What is the difference between horizontal and vertical privilege escalation?+
Horizontal escalation is reaching data or actions belonging to someone at your own level — one customer reading another customer's order, one employee opening a colleague's record. Your rank does not change; the set of things you can touch widens sideways to include peers you were never meant to reach. Vertical escalation is gaining a higher rank than you were granted — an ordinary user acquiring an administrator's powers. The reason we look for both, and for chains that combine them, is that the most serious findings are usually a horizontal step that exposes something which enables a vertical one. Each move looks minor in isolation; the path from ordinary user to administrator is what they add up to, and it is the path, not the individual step, that the fix has to close.
You say you fix the model, not the bug. What does that mean in practice?+
It means the recommendation targets the structure that produced a finding, not only the instance we traced. If an ordinary account could reach a peer's record because one endpoint failed to check ownership, the bug-level fix is to add the check to that endpoint. The model-level fix asks why the check was absent — whether object ownership is enforced consistently, or left to each endpoint to remember — and corrects that, because the same omission almost certainly recurs elsewhere. Likewise, an over-broad role is not just trimmed; we look at why roles are granted so generously and how they are created, so the next role is not another over-grant. Fixing instances leaves you patching the same class of finding at every review; correcting the model is what makes the next review shorter.
In this cluster
Next step
Map your access model with us
A review that maps the access your estate enforces, compares it against your intended policy, and returns the escalation paths and the model-level changes that close them rather than a list of one-off permission fixes.
Map your access model with us
What happens next
- 1We reply within one business day.
- 2A 30-minute scoping call with a senior consultant.
- 3A written proposal with scope, team, and timeline.
Or reach us directly

