Finance's scanner only speaks basic authentication — how do you decide what happens to that path?
answer
- Cannot prompt, so reprice instead
- Send-only, its own identity
- Whose budget replaces the device?
- Scope the allowance, not the tenant
- Undated exceptions are permanent
basics
~20 sYou cannot add a factor to a path that cannot prompt, so the decision is which control class replaces it and who pays. Scope the credential, restrict its source, or fund the device's replacement, with a named owner accepting what is left.
solid answer
~50 sFrame it as a decision about a path, not a device. A factor is structurally unavailable here, so the only levers are removing the path or repricing what the credential reaches. In practice: give the scanner its own identity rather than a person's, grant it send-only rights so the credential opens no mailbox to read, accept it only from the source addresses that device actually uses, and put a dated end on the exception. Then make ownership explicit — the business owner who needs the scanner working through year-end is the person who funds replacement or signs for the residual, not the engineer who found the path. The failure mode to avoid is both extremes: blocking legacy authentication on a Friday and breaking a finance deadline, or writing the exception into a register and letting it live for four years with a credential that opens a whole mailbox.
go deeper
Understand that some devices can only authenticate with a password, and that the answer is to limit what that password can reach rather than to add a factor the device cannot present.
Explain the substitute controls concretely: a dedicated identity, send-only rights, source restriction and a dated expiry, and why each one reduces what a stolen credential is worth.
Show the sequencing under a real deadline: the free changes now, the scoped allowance instead of a tenant-wide one, the block everywhere else, and the replacement plan with a date.
Own the negotiation. Be ready to say who holds the budget, what a written vendor commitment is worth, and how you state a residual concretely enough that accepting it is a real decision rather than a signature.
## Why this is a decision and not a fix The technical answer takes one sentence: a path that cannot present a challenge cannot be made to require one. Everything after that sentence is organisational — what the path is worth to an adversary, what it costs the business to close, and who is entitled to make the trade. That is why this question separates a senior engineer from a lead. The senior answer is "disable basic authentication". The lead's answer includes what happens when a department with a statutory reporting deadline says no. ## Establish what the credential actually reaches Before arguing about the path, price it. A scanner that submits mail typically needs to **send** and nothing else. Yet these devices are very often configured with a person's account — frequently a long-serving finance manager's — because that was the fastest way to make it work on installation day. That detail is the whole risk. The credential is not "a scanner's password", it is a working credential for a person's mailbox on a path where no factor applies. To an operator whose business is reselling access, a mailbox is a product; a send-only submission identity with no readable contents is close to worthless. **Repricing the credential is usually a larger risk reduction than anything else available, and it needs no budget** — it is a configuration change on an identity nobody owns emotionally. ## The levers, in order of what they remove 1. **Retire the path.** Replace or reconfigure the device so it authenticates modernly, or route its output through a system that does. This removes the adversary's dependency rather than raising its price. It costs money and a project slot. 2. **Separate the identity.** Its own account, never a person's, so a leaked credential cannot open anyone's correspondence. 3. **Cut the rights to the minimum.** Send-only; no mailbox read; no directory access; no ability to change its own credential. 4. **Restrict the source.** Accept the credential only from the addresses the device genuinely uses. A credential that works from one subnet is worth far less to a broker than one that works from anywhere. 5. **Date the exception.** An end date with a named owner, not an entry in a register. Undated exceptions become permanent. Notice that steps 2 to 5 are all *control classes that do not require a prompt*. That is the general principle worth stating out loud: **when the missing control cannot be reinstated, substitute controls that remove what the attack depends on** — a readable mailbox, reachability from anywhere, an unbounded lifetime. ## Who decides Three parties have standing and they are frequently confused: - **The business owner** who depends on the device and controls the budget for replacing it. They can accept the residual and they can fund the fix; they cannot overrule a hard regulatory or contractual requirement. - **The platform owner** who runs the identity provider and carries the consequences of the path being open to everyone, not just finance. Their legitimate position is that a tenant-wide protocol allowance is not a per-department decision. - **The vendor**, if the device genuinely cannot be updated. The right response is a dated statement from them — either a firmware commitment or a written confirmation that none is coming, which converts a technical annoyance into a procurement fact and usually unlocks the budget. The engineer's job is to make the choice legible: here is the path, here is what the credential reaches today, here is what it reaches after the free changes, here is what closing it costs and when. Then hand the decision to whoever owns the money. ## Sequencing under a deadline A good answer explicitly refuses both failure modes. Do not block the protocol tenant-wide the week before a close — an outage you caused is a real cost and it burns the credibility you need for the next argument. Do not accept the exception unchanged either, because the cheap changes are available today and shrink the loss by an order of magnitude. The defensible sequence is: reprice the credential this week at no cost; scope the protocol allowance to that one identity and source rather than the tenant; agree a dated replacement with the owner who funds it; block the protocol everywhere else immediately. **The exception should end up narrower than the department asked for and longer-lived than you wanted, and both of those are the point of negotiating rather than decreeing.**
- The scanner is configured with a finance manager's account. Why is that the first thing to change?Because it turns a device credential into a working credential for a person's mailbox on a path with no factor. Moving to a dedicated send-only identity costs nothing, needs no project, and removes the readable contents that make the credential saleable. It also stops the manager's password rotation from breaking the scanner, which is why it was set up that way in the first place.
- The vendor says modern authentication is 'on the roadmap'. What do you do with that?Ask for a dated commitment in writing. A date converts the problem into a procurement fact you can plan against; a refusal or a vague answer is equally useful, because it justifies the replacement budget. An undated roadmap promise is the mechanism by which a temporary exception becomes a four-year one.
- Who signs off leaving the path open, and what exactly are they accepting?The business owner who funds the replacement, not the engineer who found it. They are accepting a specific, priced outcome: that one credential, if obtained, permits sending as this identity from these sources until this date. If you cannot state it that concretely, the acceptance is theatre and nobody has actually decided anything.
saying these in an interview costs you the question
- Blocks the protocol tenant-wide without asking what breaks
- Accepts the exception without repricing the credential
- Leaves the device on a person's account
- Writes an exception with no end date and no owner
- Treats the engineer who found it as the person who signs it off