About this article: This article is reviewed against the cited sources. This is source research, not lab validation. Check the stated versions and your own requirements before making changes.
Your hosting or backup provider takes requests: restore a server, recover a folder, reset a password, add a user, send a copy of the backups. Most come from you. Some may not. If a provider acts on a request that only sounds like it came from you, the damage can be a deleted system, or a copy of your data in someone else's hands.
The short answer: agree in writing, before you need it, who may make each kind of request, through which channel, and how the provider will confirm it. The sensible minimum is a named list of authorised requesters, a rule that the provider confirms by calling back a number already on file (never one supplied in the request), and a second approver for the riskiest actions. The business sets the authority. The provider applies it.
Why it matters
Impersonating a trusted person to a help desk is a documented method. A joint advisory by agencies including the US CISA and FBI and the Australian Signals Directorate's Australian Cyber Security Centre describes the Scattered Spider group. Its technique table says its members "use voice communications to convince IT help desk personnel to reset passwords and/or MFA tokens" (CISA AA23-320A, read through an archive copy). It also lists trusted relationships of contracted IT help desks as a way in, and says the group gathers personal information from open sources to make its calls convincing.
That advisory is about help desks and account resets, not restores. Applying its lesson to backup and restore requests is our reading. The reasoning is simple: a provider that holds your backups has people who can restore or hand over data when asked, and that is a strong lever for an impersonator.
The Australian Signals Directorate's Small business cyber security handbook gives the basic defence against business email compromise, where criminals impersonate a supplier, customer or staff member. Its advice is to "Verify unusual requests using a trusted communication channel, such as a known phone number rather than contact details provided in the email." (ASD, Small business cyber security handbook, Common cyber threats to small businesses, published and last reviewed 7 October 2026). That is email-fraud guidance for small businesses. It is not a standard for restore requests. Our recommendation is to apply the same known-number rule to requests that reach your provider, including restores.
Different requests carry different risks
Backup, replication, failover and disaster recovery are not the same thing, and the requests differ. A restore returns data from a backup copy. A replica or failover site is a running copy that takes over. For this article we mean requests to restore or release data and to change who may ask. The following grouping is our recommendation, not a standard.
| Request | What could go wrong if forged | Suggested check |
|---|---|---|
| Restore a file or folder to its original place | Overwrites current data with an older version | Authorised requester, confirmed by call-back |
| Restore a whole system over the live one | Loss of recent work, outage | Authorised requester plus a second approver |
| Restore to a new or different location | A copy of your data lands somewhere an attacker chooses | Second approver, and the target named in advance |
| Export, download or send backup data | Data leaves your control | Second approver, written destination |
| Reset a password or multifactor method | Account takeover | Strongest check, never by email alone |
| Delete backups or shorten retention | Removes your ability to recover | Second approver and a waiting period |
| Add, remove or change an authorised requester | Replaces the safeguard itself | Strictest check, confirmed with every existing contact |
The last row matters most. If an impersonator can add themselves to the list, every later check passes.
What makes a request genuine
A request is genuine only if four things hold. The person is authorised for that kind of request. The channel is one you agreed to. The scope (which system, which restore point, which destination) matches what the business intends. And, for the riskier kinds, a second person has approved it.
The things that feel like proof are weak. The advisory describes callers who already know personal details, because they collected them beforehand. A familiar voice, a correct employee name, a plausible urgent reason and a matching email display name can all be faked or gathered. Caller ID can too, so a call alone does not prove a caller is who they claim to be (our reading).
The Information Security Manual (ISM), the Australian Signals Directorate's security framework, says personnel dealing with suspicious requests to change user account details "should have clear guidance to think critically before actioning unusual requests as well as a process to report threatening demands for immediate action, pressure for secrecy or requests to circumvent normal business processes and procedures" (ISM, Guidelines for personnel security). It also has a control (ISM-2071, updated September 2025) that such personnel are told what social engineering attacks are, how to manage them and how to report them. The ASD describes the ISM as a cyber security framework that an organisation can use. For a private business it is guidance, unless a law, a regulator or a contract requires particular controls from it. Whether that applies to you is a question for your own advisers. Its wording is a useful model for both sides.
The call-back rule
Our recommendation is a simple rule for every request that is not routine:
- The provider does not act on the first contact.
- The provider ends the contact and calls back a number already recorded for that authorised person, taken from the agreed list.
- The caller confirms the request, and a reference from the provider's ticket, in that call.
- For the riskier requests, the provider also asks the second approver.
- The provider records who asked, who approved, and what was done.
The number must come from the list, not from the request. If a message asks you to ring a new mobile number, the answer is to use the old number on file.
The rule has limits. The advisory says the group also uses SIM swap attacks, which move a phone number to a device the attacker controls. A call-back to a mobile number is therefore not perfect, which is why the riskiest actions also need a second approver and why changes to the list need the strictest check. The advisory recommends phishing-resistant multifactor authentication (FIDO or WebAuthn, or certificate-based), which it says is not susceptible to push bombing or SIM swap attacks. Our reading: for requests made through a provider's online portal, ask whether the portal supports it.
What to agree in writing
Our recommendation for the contract, service description or an agreed procedure:
- Authorised requesters. Names, roles, and for each person the kinds of request they may make. Include at least two people with authority over the highest-risk actions, so one absence or one compromised person does not stop you or expose you.
- Second approver. Who confirms the high-risk requests, and that this must be a different person.
- Channels. Which channels count: portal with strong sign-in, phone call-back, or a ticket from an agreed address. Which do not count on their own, such as a bare email.
- Pre-agreed destinations. Where restored or exported data may go, so a new destination triggers extra checks.
- Out-of-hours path. An emergency route that still includes verification, since urgency is exactly what a fraudster will claim.
- List changes. How a person is added or removed, and that all current contacts are told.
- Records. What the provider logs and how you can see it.
- Reporting. That the provider tells you promptly about a refused or suspicious request.
- Review. A check of the list after each staff change and at least once a year.
The ISM's guidance on procurement and outsourcing is a useful pointer for what an organisation keeps about a provider. Its registers for managed and outsourced cloud services include the provider's name, an organisational point of contact and 24/7 contact details for the provider (ISM, Guidelines for procurement and outsourcing, controls ISM-1737 and ISM-1638). Our reading: keep the same two-way contact record, so you know who to call at the provider as well as who the provider may call.
Read-only checks
None of these change a system. They review documents and past activity.
- Find the authorised contact list your provider holds for you. Does it match who works for you today?
- Look at the last three restore, access or reset requests. For each, who made it and how was it confirmed?
- Check whether any person on the list has left, changed role, or shares a mailbox with others.
- Check where the provider sends notifications about requests, and that the address is monitored.
- Check what your contract or service description says about verification. If it is silent, treat that as a gap to close.
Questions to put to your provider
- How do you verify that a restore, access or reset request is genuine, for each channel you accept?
- Do you ever act on email alone, or on a call you did not originate?
- How is the list of authorised requesters changed, and who is told?
- Is a second approver required for restores to a new location, exports and deletions?
- Can we set pre-agreed destinations for restores and exports?
- What do you log, and can we see it?
- What happens out of hours?
A drill is a good test (our recommendation), but do not use real data. The authorised-contact list is itself useful to an impersonator, so do not ask the provider to read it out or send it. Agree the drill with the provider beforehand, in writing, using a made-up request with dummy names and details, and no real restore, export or reset. Then see whether the provider's verification steps match what you agreed.
A worked example (illustrative)
The following is invented. A firm's provider receives an email, apparently from the firm's finance manager, asking urgently for the last night's backup of the accounting server to be restored to a new cloud storage location, saying it is for an audit. The email is polite, uses correct names and mentions an invoice number from a real supplier.
Under the firm's agreed procedure, a restore to a new location needs a call-back to the requester on the number on file, plus the second approver, and a destination named in advance. The provider calls the finance manager, who knows nothing of it. The provider declines, logs the attempt and tells the firm's named security contact. The firm changes its email passwords, and the provider notes the incident in the request record.
Read next
Related articles you may want next.
- Backup & RecoveryQuestions to ask a managed backup provider before you signQuestions to ask a managed backup provider about restores, access, incident notice, data location and exit, drawn from ASD outsourcing guidance.Read article
- Backup & RecoveryA minimum recovery plan for a small business with no IT staffA short backup and recovery plan a small business owner can write without IT staff, based on ASD's 2026 small business cyber security handbook.Read article
- Backup & RecoveryOut-of-band communications when identity and email are downHow to reach staff, leaders, suppliers, and customers when email and single sign-on are down or unsafe during ransomware recovery.Read article
Risks and limits
- No verification method is perfect. SIM swaps, compromised email accounts and insiders can defeat parts of it.
- The advisory and the ISM guidance were written for other audiences, such as large organisations and government. We apply them to small business by analogy, and we have not found a source that sets a verification standard for restore requests specifically.
- Stricter checks cost time. Agree a clear emergency path so people do not bypass the rule.
- The right design depends on your contract, your provider's tools and your risk. These are general suggestions, not a design for your business.
Next step
Write down, this week, the named people who may ask your provider for restores and resets, the second approver for high-risk requests, and a rule that the provider confirms by calling a number already on file. If you are reviewing a backup service, the Backup and Recovery page is a starting point for that conversation; ask for the scope in writing.
Sources
- Scattered Spider (AA23-320A), CISA with FBI, RCMP, ASD's ACSC, AFP, CCCS and NCSC-UK, first published 16 November 2023, updated 29 July 2025 (read through an Internet Archive copy captured 10 November 2026, because the CISA site refused automated access).
- Guidelines for personnel security, Australian Signals Directorate, Information Security Manual, September 2026 release (read at the current address on 16 November 2026).
- Guidelines for procurement and outsourcing, Australian Signals Directorate, Information Security Manual, September 2026 release (read at the current address on 16 November 2026).
- Information security manual, Australian Signals Directorate, September 2026 release (read at the current address on 16 November 2026).
- Small business cyber security handbook: Common cyber threats to small businesses, Australian Signals Directorate, published and last reviewed 7 October 2026 (read at the current address on 16 November 2026).
All accessed 15 November 2026, except where a source line gives 16 November 2026.
Learning notes
- A restore request is powerful, so treat it like a payment request: confirm it before acting.
- Impersonating staff to a help desk is a documented technique, and phone calls and emails can be faked.
- Confirm by calling a number already on file, never one given in the request.
- Riskiest requests: restore to a new place, export, delete, reset sign-in, and change the authorised list. Give them a second approver.
- Agree the list, channels and rules in writing, and review after staff changes.
This article was originally published on .