skip to content

One scanner fleet keeps SMBv1 alive for 600 hosts - how do you actually retire the protocol?

level: principalimportance: nice to knowfreq 33%

answer

  1. reframe 600 hosts as one device
  2. check firmware before costing hardware
  3. capital spend, not a security change
  4. an owner who is allowed to refuse
  5. narrow the exception, expire it

basics

~20 s

Treat it as a capital replacement, not a security change. Find the one remaining consumer, upgrade or replace it against the budget that owns the device, attach an owner and a date, and scope the exception to that device meanwhile.

solid answer

~50 s

Retirement here is a purchasing decision wearing a security label. The sequence is: establish which consumer is actually holding the protocol open - usually one device family, not six hundred hosts; confirm whether a firmware release removes the requirement, because that is far cheaper than replacement; if not, cost the replacement and take it to the budget that owns those devices, which is rarely security. Until the money lands, narrow the exception from a fleet-wide permissive setting to the one device that needs it, cut what that device can reach to what it must, and attach a named owner and a date to the exception so that someone can be asked to renew or refuse it. Do not accept 'accepted risk, no expiry'. And be honest that a security team answering 'just turn it off' loses the argument the first time a scanner stops working, after which a vendor engineer quietly turns it back on.

go deeper

for a junior

Know that a legacy protocol usually survives because one device still needs it, and that finding that device comes before changing anything fleet-wide.

for a middle

Explain why narrowing the exception to one host and uninstalling elsewhere is stronger than leaving a permissive setting everywhere with a note attached.

for a senior

Show the operational sequence - identify the consumer, check firmware, scope the exception, cut what the device can reach - and anticipate the vendor engineer undoing your setting.

for a principal

Own the funding and accountability call: name the budget that pays, get an owner who is permitted to refuse, attach a date, and be willing to carry an explicit refusal rather than an anonymous exception.

## Why this is not a technical question Everything technical about retiring a legacy protocol is easy. The setting is one line, the removal is one optional feature, and both are well documented. What makes this a lead's question is that the last consumer is a physical object owned by somebody who did not ask for a hardening programme and has no budget line for it. So the answer has to be shaped like a purchase, not like a change ticket. ## Step one: find the single consumer The framing '600 hosts still allow SMBv1' is almost always wrong, and challenging it is the first thing a strong answer does. Six hundred hosts do not each independently need the protocol. One device family does - a multifunction scanner that writes scans to a share, a badge controller, a camera recorder - and the fleet-wide permissive setting exists because somebody once needed the scanner to work and the fastest fix was the broadest one. The question to answer is therefore: **which single device is holding this open, and for how many hosts?** That reframes a 600-host hardening problem into a one-device procurement problem, and it is the reframe the interviewer is listening for. ## Step two: check the cheap exit first Before costing hardware, establish whether the vendor has a firmware release that speaks something modern. Appliance vendors frequently do, several years late, and the upgrade is a fraction of a replacement. This step is skipped surprisingly often because the device is treated as an immovable object rather than as a supported product with a release stream. ## Step three: cost it as capital, and give it an owner who can refuse If the device cannot be upgraded, retirement means replacing it. That is capital spend against the budget that owns the device - facilities, print services, physical security - not against the security budget, which usually cannot buy scanners. The output of this step is a named owner, a number, and a date. The named owner matters more than the date, and this is the part inexperienced answers miss: the owner must be someone who is allowed to **refuse**. An exception nobody can refuse is not a decision, it is a drift. If the print services owner looks at the number and says 'not this financial year', that is a legitimate outcome, and now the risk is held by someone who can actually trade it against other spending. What is not legitimate is the exception existing with no name attached. ## Step four: narrow the exception while you wait Almost always the money does not land immediately, so the interim matters. The move is to shrink the exception from the estate to the device: - the permissive setting applies to the one host or the one device, not fleet-wide; - everywhere else the feature is uninstalled rather than disabled, so the exception cannot silently re-spread through a policy scope change; - the device's reachability is cut to only the share and the hosts it genuinely needs; - the exception carries an expiry, so it comes back for a decision instead of becoming permanent. On a flat office segment this last point is doing real work, because the neighbours of that scanner are badge controllers and cameras, and an adversary who wants publicity rather than depth is perfectly happy with a camera feed as the objective. 'It is only a printer' undervalues both the device and the company it keeps. ## Step five: expect the setting to be undone The reason the interim must be a narrowed *removal* elsewhere rather than a broad *disable* is behavioural. The vendor's field engineer holds delegated administration so they can service the fleet. When a device breaks after a hardening change, they will restore whatever makes it work, without malice and without telling anyone. A disabled dialect survives that for exactly as long as nobody needs the scanner. An uninstalled component does not silently come back. ## What a weak answer sounds like 'We disable it fleet-wide and grant an exception for the scanner.' That is the same permissive state described more politely, it has no owner, no date and no purchase behind it, and in two years it will still be there with everyone believing the protocol was retired. The whole point of the disabled-versus-removed distinction is that only one of those two states is a claim you can still defend after somebody else has had administrative rights for two years.

  • The device owner refuses the spend this year. Have you failed?
    No, provided the refusal is explicit, named and dated. A refusal by someone who can weigh it against other spending is a decision; the failure mode is an exception that exists because nobody was ever asked. Your remaining job is to keep the exception scoped to that device and bring it back at the next cycle.
  • Why not just leave the fleet-wide setting permissive until the replacement lands?
    Because a broad permissive setting is indistinguishable from never having started, and it re-spreads. Uninstalling the component everywhere except the one device means a future policy scope change or a vendor engineer's fix cannot quietly restore it across the estate, only on the one host that already carries the exception.
  • It is only a scanner on an office segment - why treat it as significant?
    Because of what it sits next to. A flat site segment carries badge controllers and cameras alongside it, and an adversary optimising for publicity needs no depth at all when a camera feed is itself the objective. The device's own value is a bad proxy for what reaching it is worth.

saying these in an interview costs you the question

  • Proposes disabling fleet-wide and calling the protocol retired
  • Assumes the security budget can replace facilities hardware
  • Leaves the exception with no owner and no expiry
  • Never asks which single device holds the protocol open
  • Skips checking whether a firmware release removes the need

context