Storage media leave the organisation every week, through warranty swaps, leasing returns, cloud volume deletion and recycling, and each one is a data breach waiting to happen if the data on it can still be recovered. NIST SP 800-88 Rev. 2, Guidelines for Media Sanitization, was published as final in September 2025 and replaces Rev. 1 from December 2014. The revision moves away from the per-device tables that most teams knew and towards a sanitisation programme: policy, decision criteria, assurance and records, with technology-specific techniques delegated to IEEE 2883. This guide explains what changed, how to choose between clear, purge and destroy, when cryptographic erase can be trusted, and what evidence an auditor should expect. It is written for infrastructure, asset management and security teams that own decommissioning, and for assessors who test it.
Key takeaways
- Rev. 2 is a programme standard, not a lookup table. The old media-by-media appendix is gone; NIST now expects a policy, a decision framework, verification and validation, and documentation, and points to IEEE 2883 for the technique on each device type.
- Purge is the default, clear is the exception. NIST states that purge should be used instead of clear when possible, and clear protects only against simple, non-invasive recovery through the normal interface.
- Cryptographic erase is only as good as the keys. It is a fast, high-assurance purge technique when no plaintext was ever written, every key copy can be sanitised and no escrowed or backed-up key survives outside the device.
- Flash breaks magnetic-era habits. Multi-pass overwriting gives very little protection on SSDs with overprovisioning, degaussing does not work on flash, and pulverising or shredding is acceptable only for the lowest categories of data.
- Verification and validation are separate steps. Verification checks that the technique completed; validation is the decision to accept or reject the sanitisation, and both need a named person and a certificate of sanitisation.
What it is and who it applies to
SP 800-88 defines media sanitisation as a process that renders access to target data on the media infeasible for a given level of effort. That wording matters: sanitisation is judged against an attacker's effort and the sensitivity of the data, not against an absolute standard of "wiped". The publication covers two media classes: hard copy (paper, film, microfiche) and information storage media (ISM), which includes magnetic disks, flash (SSD, NVMe, USB, SD, eMMC), optical discs, the storage inside phones, printers, network appliances and firewalls, and logical or virtual storage such as cloud volumes.
- Federal agencies: SP 800-88 is the reference that SP 800-53 Rev. 5 control MP-6 Media Sanitization points to, so it is in scope for FISMA systems, FedRAMP cloud services and contractors that inherit those controls.
- Everyone else: outside government it is the most widely cited benchmark for sanitisation. IT asset disposition (ITAD) contracts, cyber insurance questionnaires and auditors commonly ask whether disposal "meets NIST 800-88", and ISO/IEC 27001 certification audits look for an equivalent documented method.
- What it does not cover: Rev. 2 states that technology-specific sanitisation techniques are out of scope, with the exception of cryptographic erase. Those live in IEEE 2883-2022, Standard for Sanitizing Storage, and its companion recommended practice IEEE 2883.1, plus NSA/CSS Policy Manual 9-12 for classified contexts and DISA STIGs.
For the wider records and retention lifecycle that decides when data should be deleted at all, see the Data Retention and Secure Deletion Guide. For the commercial side of disposal providers, see the IT Asset Disposition Security Guide.
Structure and key requirements
Rev. 2 is organised around three sanitisation methods and a programme that decides which one applies. The methods are defined by outcome; the techniques that achieve them depend on the device.
| Method | Outcome NIST describes | Typical techniques | Media after |
|---|---|---|---|
| Clear | Protects against simple, non-invasive recovery through the normal user interface. Never appropriate for hard copy. | Logical overwrite with standard read and write commands, factory reset | Reusable |
| Purge | Recovery infeasible even with state-of-the-art laboratory techniques, while leaving media potentially reusable | Logical: overwrite, block erase, cryptographic erase using device sanitise commands. Physical: degaussing (magnetic media only) | Usually reusable (degaussed drives are not) |
| Destroy | Recovery infeasible with state-of-the-art laboratory techniques and media unusable for storing data | Disintegrate, incinerate, melt, pulverise, shred | Unusable |
The programme elements
Section 4 sets out what a media sanitisation programme should contain. Drawing on ISO/IEC 27040, NIST lists policy, scope, a decision framework, performing sanitisation, assurance and documentation, supported by defined roles.
- Policy (4.1): map the data classification scheme (low, medium, high) to minimum methods (clear, purge, destroy); define disposal and reuse expectations and acceptable techniques by reference to IEEE 2883; set competency, training and tool calibration requirements; and state what assurances storage vendors must provide.
- Scope (4.2): NIST distinguishes selective, partial and whole-device sanitisation and prefers whole-device sanitisation where possible.
- Decision framework (4.3): the decision is driven by the confidentiality of the information, not the media type. The questions are whether the media will be reused, whether it will leave organisational control, and what protection level the data needs. Media that will not be reused can often simply be destroyed, which NIST notes may be the simplest and most cost-effective option.
- Organisational control (4.3.4): media under a maintenance contract that protects confidentiality, or maintained on site under supervision, can stay "in control". Warranty exchanges, rebates and resale are typically outside control, and losing control of unsanitised media may be treated as a data breach under legal or regulatory rules.
- Assurance (4.5): verification of the technique's outcome, then validation as an approve or reject decision.
- Documentation (4.6) and roles (4.7): a certificate of sanitisation (sample in Appendix C) and ten named roles from programme managers to users, including property management, records management and privacy officers.
Choosing a method
| Situation | Practical minimum | Why |
|---|---|---|
| Low-impact data, device redeployed internally | Clear (or purge if the device supports a sanitise command) | Stays inside organisational control; purge is preferred where cheap |
| Moderate or high data, device leaving control for resale, lease return or donation | Purge with a device sanitise command or verified cryptographic erase | Laboratory recovery is the realistic threat once media leaves |
| High data, device not reused, or purge cannot be verified | Destroy, with particle size set by policy | Removes reliance on firmware behaving as claimed |
| Failed drive under warranty | Retain the drive (non-return option) or destroy; do not ship unsanitised | A drive that cannot accept commands cannot be purged logically |
| Cloud volumes, snapshots, object storage | Cryptographic erase through key deletion, backed by provider attestations | Physical media are not accessible; the key is the only lever |
| Hard copy | Destroy | Clear is never appropriate for hard copy |
Cryptographic erase conditions
Section 3.2 treats cryptographic erase (CE) in more depth than any other technique, because it is the one technique that is not device specific. CE sanitises the keys that encrypt the data, or prevents access to them. NIST attaches conditions:
- Strength: the cipher and mode should provide at least 128-bit security strength, with random sources whose entropy at least equals the key length. ECB mode is not acceptable, and federal agencies must use modules validated to the current FIPS 140 standard.
- No prior plaintext: CE works only if no sensitive data was written to the media unencrypted since its last sanitisation. A laptop that ran unencrypted before BitLocker was enabled is not a CE candidate.
- Every key copy: all copies of target keys must be sanitisable, keys lower in the hierarchy must be eliminated, and keys already unwrapped into memory or registers must be cleared, for example by a hard reset or power off. Keys held outside the device, such as in a key management system, are a recovery risk.
- Backups and escrow: NIST says CE should not be trusted on media that has been backed up or escrowed unless off-device key storage is well controlled.
- Long-lived data: because future cryptanalysis or quantum computing could make ciphertext readable, CE may not be acceptable for data that must stay confidential for a long time.
The GoTo incident shows why the key-copy rule matters: the GoTo breach involved stolen encrypted backups together with an encryption key for part of them, so the encryption offered no protection for that portion. A key that survives somewhere else undoes cryptographic erase in exactly the same way.
Identity and NHI requirements
SP 800-88 is about data, but decommissioned media are also one of the largest unmanaged stores of machine credentials. A server disk, a firewall's flash or a developer laptop typically holds service account passwords, SSH private keys, API tokens in configuration files, Kerberos keytabs, TLS private keys and cached cloud credentials. Sanitising the media removes the copy on that device; it does nothing about the same credential still being valid elsewhere.
- Revoke before you wipe: make credential revocation and rotation a gate in the decommissioning workflow. Disable the computer account, revoke device certificates, delete or rotate any service account and API key the host used, and remove the host's SSH key from authorized_keys files. See the NHI Offboarding and Decommissioning Guide.
- Treat media encryption keys as identities: CE depends on knowing where every data encryption key, key-wrapping key and recovery key lives. BitLocker recovery keys escrowed to Entra ID or Active Directory, FileVault institutional keys and KMS keys for cloud volumes are all copies that must be deleted or deliberately retained. See the Cryptographic Key Management Guide.
- Hardware roots of trust: TPM-held keys and attestation identities should be cleared (TPM clear) before redeployment, so a new user does not inherit device identity or sealed secrets.
- Access to sanitisation tools: management consoles that can issue remote wipe or sanitise commands (MDM, out-of-band controllers, storage array managers) are privileged. Restrict them, log their use and require approval, because the same capability can destroy production data.
Implementation steps
- Inventory media, not just assets. Extend the CMDB so each asset lists its storage components: internal drives, embedded flash, removable cards, and the storage inside printers, copiers and network appliances. Appendix B lists the device characteristics to request from vendors, and a Statement of Volatility identifies which components retain data. See the Asset Inventory and CMDB Security Guide.
- Write the policy matrix. Map each data classification to a minimum method, and each media type to approved techniques by reference to IEEE 2883. Decide up front whether failed drives are ever returned under warranty.
- Standardise encryption at deployment. Encrypt every device from first boot (self-encrypting drives, BitLocker, FileVault, LUKS) so CE becomes available later. Record whether the device ever held plaintext. See the Disk Encryption Guide.
- Select and qualify tools. Prefer firmware sanitise commands (ATA Sanitize, NVMe Sanitize and Format with secure erase settings, SCSI SANITIZE) or vendor tools that issue them, over host-based overwriting of flash. Record tool name and version; NIST expects calibration, testing and maintenance of equipment.
- Pilot on each media family. Run the technique on sample devices of each model, then verify with a forensic read of accessible LBAs and the tool's completion status. Confirm the drive reports success and that reuse configuration (re-partitioning, firmware state) still works, since NIST notes sanitised drives will not look factory fresh.
- Build the decommissioning workflow. Order: credential revocation, data owner and records officer sign-off (to avoid destroying records under legal hold), sanitisation, verification, validation, certificate, asset status update.
- Contract the outside world. ITAD providers, maintenance contractors and cloud providers must state their methods, provide per-serial certificates and allow audit. Chain of custody covers transport.
- Plan the fallback. If verification fails or a drive will not accept commands, the policy escalates to the next method (usually destroy). That escalation is the "rollback" path; nothing leaves control until validation approves it.
Evidence and assessment
An assessor testing MP-6 or ISO/IEC 27001 Annex A 7.14 will sample disposed assets and trace each one to a record. Appendix C's sample certificate gives the minimum fields.
| Evidence | What it should show | Common gap |
|---|---|---|
| Certificate of sanitisation | Manufacturer, model, serial, property number, media type and source, method and technique, tool and version, verification method, verifier and validator name, title, date, location, signature | Batch certificates with no serial numbers |
| CE traceability record | Make, model, key generation, media encryption, key wrapping, areas covered, key lifecycle, key sanitisation technique, escrow or injection, error handling | No record of whether recovery keys were escrowed |
| Asset register reconciliation | Every retired asset has a sanitisation record or a documented exception | Drives pulled from servers never entered the register |
| Tool qualification | Pilot results per media family, tool versions, calibration for degaussers and shredders | Overwrite tool used on NVMe without testing |
| Third-party evidence | ITAD certificates per serial, chain of custody, right to audit | Weight-based recycling receipts only |
Metrics that show it works:
- Percentage of retired assets with a per-serial certificate within 30 days of retirement (target 100%).
- Number of media leaving control without validation (target zero; each one is an incident).
- Percentage of fleet encrypted from first use and therefore eligible for CE.
- Verification failure rate by media model, which shows when a technique or firmware is unreliable.
- Mean time from retirement to sanitisation, since drives waiting in a cupboard are the most common loss point.
- Percentage of decommissioned hosts with machine credentials revoked before sanitisation.
Mapping to other frameworks
| Framework | Reference | Relationship to SP 800-88 |
|---|---|---|
| NIST SP 800-53 Rev. 5 | MP-6 Media Sanitization; MP-6(1) Review, Approve, Track, Document, and Verify; MP-6(2) Equipment Testing; MP-6(3) Nondestructive Techniques; MP-6(8) Remote Purging or Wiping of Information | SP 800-88 is the method guidance behind these controls |
| NIST SP 800-53 Rev. 5 | SR-12 Component Disposal; MP-4 Media Storage; SC-28 Protection of Information at Rest; SC-12 Cryptographic Key Establishment and Management | Disposal in the supply chain, storage pending sanitisation, encryption that enables CE, key destruction |
| NIST CSF 2.0 | ID.AM-08 lifecycle management; PR.PS-03 hardware maintained, replaced and removed commensurate with risk; PR.DS-01 data at rest protected | Outcome level statements the programme satisfies |
| ISO/IEC 27001:2022 Annex A | 7.10 Storage media; 7.14 Secure disposal or re-use of equipment; 8.10 Information deletion | Certification audits expect a documented method and records |
| CIS Controls v8.1 | Safeguard 3.5 Securely Dispose of Data | Requires a documented, classification-driven disposal process |
| IEEE 2883-2022 | Standard for Sanitizing Storage | Technology-specific clear, purge and destroy techniques that Rev. 2 now delegates |
| MITRE ATT&CK | T1005 Data from Local System; T1052 Exfiltration Over Physical Medium | Techniques that unsanitised media make trivial once a device is outside control |
Common pitfalls
- Running a 3-pass or 7-pass overwrite on SSDs. NIST warns that multi-pass overwriting from the magnetic era gives very little protection on overprovisioned flash, where wear levelling leaves stale copies in spare cells the host cannot address. Use the device's sanitise command or CE.
- Trusting "encrypted, so safe" without checking keys. A self-encrypting drive with the default (unset) authentication, a BitLocker volume whose recovery key is still in the directory, or a cloud volume whose KMS key is shared with other volumes has not been cryptographically erased. Key copies must be found and destroyed.
- Degaussing flash and calling it destroyed. Degaussing has no effect on non-magnetic media and NIST does not treat it as an approved destroy technique.
- Forgetting non-obvious storage. Multifunction printers, VoIP handsets, firewalls, load balancers and vehicle telematics hold credentials and data. Factory reset is usually only a clear.
- Shipping failed drives back to the vendor. A drive that will not accept commands cannot be purged; a warranty return of an unsanitised drive is a loss of control. Buy non-return options for drives holding moderate or high data.
- Certificates without serial numbers. A certificate that says "200 drives destroyed" cannot be reconciled to the asset register, so a missing drive is never noticed.
Practitioner checklist
- Update the sanitisation policy to Rev. 2: classification to method matrix, IEEE 2883 references for techniques, roles from section 4.7.
- Inventory storage components per asset, including embedded flash in appliances and printers.
- Encrypt every device from first use and record whether plaintext was ever written, so CE is defensible.
- Locate every copy of media encryption and recovery keys (directory escrow, KMS, backups) and include their deletion in the CE procedure.
- Replace host-based overwriting of flash with firmware sanitise commands or validated vendor tools, and pilot each media family.
- Add credential revocation (computer accounts, certificates, SSH keys, API tokens) as a mandatory step before sanitisation.
- Separate verification and validation, with named people, and issue a per-serial certificate with the Appendix C fields.
- Decide the warranty policy for failed drives and buy non-return options where needed.
- Require per-serial certificates, chain of custody and audit rights from ITAD and maintenance providers.
- Track the six metrics above and reconcile retired assets to certificates every month.
Standards and references
- NIST: SP 800-88 Rev. 2, Guidelines for Media Sanitization
- NIST: SP 800-88 Rev. 2 full text (PDF)
- IEEE: 2883-2022 Standard for Sanitizing Storage
- NIST: SP 800-53 Rev. 5, Security and Privacy Controls (MP-6, SR-12)
- NIST: Cybersecurity Framework 2.0
- MITRE ATT&CK: Exfiltration Over Physical Medium
Related NHI Mgmt Group resources: Data Retention and Secure Deletion Guide · IT Asset Disposition Security Guide · Disk Encryption Guide · Cryptographic Key Management Guide · NHI Offboarding and Decommissioning Guide