CCSP Study Guide — 2026 Edition

Data is the asset cloud security exists to protect. This chapter follows data through its lifecycle and covers the technologies that protect it: encryption, tokenization, hashing, key management, data discovery and classification, IRM, and retention, deletion and archiving.

2.1 Describe cloud data concepts

The CSA Data Security Lifecycle has six phases from creation to destruction: Create (generation of new digital content, or alteration of existing content); Store (committing data to a storage repository, typically nearly simultaneous with creation); Use (data is viewed, processed or otherwise used, not including modification); Share (information is made accessible to others, such as users, customers and partners); Archive (data leaves active use and enters long-term storage); Destroy (data is permanently destroyed using physical or digital means, for example cryptoshredding). Although shown as a linear progression, data can bounce between phases without restriction and may not pass through all stages.

Data dispersion — fragmenting data across multiple storage locations — reduces the value of any single compromised store. Data flows must be mapped (where data moves, in which phases, and how it might be accessed) so controls match reality.

2.2 Design and implement cloud data storage architectures

Cloud storage types include long-term storage, ephemeral storage (short-lived, vanishing when the instance or session ends), raw storage, object storage and volume storage. Each has its own threat profile: ephemeral storage can silently drop data the organization assumed was durable; object stores' flat namespaces and rich metadata change how access control is evaluated; and shared underlying infrastructure exposes all of them to multi-tenant and provider-side threats. Design storage so that the protection (encryption, access control, redundancy, sanitization) matches each type's threat model.

2.3 Design and apply data security technologies and strategies

Encryption and tokenization are not interchangeable. A token replaces sensitive data with a non-sensitive placeholder and keeps no mathematical relationship to the original: it cannot be reversed by computation, only looked up in the token vault. Encryption is reversible and maintains a mathematical relationship to the original data, so ciphertext can be decrypted by anyone who obtains the key.^[1] Data tokenization substitutes a sensitive data element with a non-sensitive placeholder; tokens are generated randomly and lack any inherent or exploitable meaning or value, functioning as a reference to the sensitive data. A token vault securely stores the relationship between tokens and the original data — protecting that vault is critical, because it holds the key to retrieving the sensitive information. Tokenization can reduce PCI DSS scope by shrinking the number of system components in scope, though it does not eliminate compliance obligations. Format-preserving tokens keep the expected shape of the data (for example, a sixteen-digit string), easing integration with applications that expect specific formats; non-format-preserving tokens do not resemble the original structure.

Key management is where most encryption programs succeed or fail. FIPS 140-3 (Security Requirements for Cryptographic Modules) superseded FIPS 140-2 for new submissions as of April 1, 2022, when NIST's CMVP stopped accepting FIPS 140-2 submissions for new validation certificates; existing FIPS 140-2 certificates move to the Historical List in September 2026^[2]; the Cryptographic Module Validation Program, a joint effort between NIST and the CCCS, validates modules, with testing done by NVLAP-accredited laboratories. FIPS defines four increasing security levels: Level 1 requires production-grade equipment and externally tested algorithms; Level 2 adds physical tamper-evidence and role-based authentication^[3]; Level 3 adds physical tamper-resistance and identity-based authentication, with private keys entering or leaving only in encrypted form^[3]; Level 4 requires tamper-active behavior (erasing contents on detecting environmental attack), fault-injection protection and multi-factor authentication^[3].

Hashing provides integrity, not confidentiality. FIPS 180-4's secure hash algorithms (SHA-224/256/384/512 and others) are one-way functions producing a condensed representation called a message digest; any change to the message will, with very high probability, produce a different digest. That property is useful in the generation and verification of digital signatures and message authentication codes: a verification failure signals the message was altered, supporting integrity and non-repudiation.

Data Loss Prevention (DLP) and access control are the core technologies for keeping confidential data from leaving company-controlled systems; CASB data-security functions rely on both.

Data obfuscation techniques such as masking and anonymization reduce exposure by replacing sensitive values with fictitious but plausible ones — useful in test environments and analytics where real values are unnecessary.

2.4 Implement data discovery

You cannot protect data you cannot find. Data discovery scans structured, unstructured and semi-structured stores to locate sensitive data and determine its location. Discovery exists precisely because much data is never classified at creation — it must be found before it can be protected, and its location recorded so controls match reality.

2.5 Plan and implement data classification

Classification identifies and labels sensitivity; a workable scheme needs at least two distinguishable levels, with labels (data labeling and tagging) that tell handlers how to treat the data. Data mapping records where each class lives and flows. Classification is cheapest and most effective when applied at creation — the first lifecycle phase — because retroactive classification depends on discovery finding the data first.

2.6 Design and implement Information Rights Management (IRM)

IRM applies persistent protection — encryption plus authentication plus use rights — to sensitive information such as email and documents. Classification, labels and protection travel with the data, so it stays identifiable and protected at all times, regardless of where it is stored or with whom it is shared; encryption settings remain with the data even when it leaves the organization's boundaries. IRM limits who can copy, print or forward content, and includes issuing and revoking certificates that carry the usage rights.

2.7 Plan and implement data retention, deletion and archiving policies

Retention policies define how long each data class is kept; deletion procedures and mechanisms destroy data when retention expires; archiving moves inactive data to long-term storage. Media sanitization — rendering access to target data infeasible for a given level of effort — implements destruction. NIST SP 800-88 defines three sanitization actions: Clear uses standard rewriting techniques, giving moderate protection against simple noninvasive recovery, and media can be reused; Purge uses state-of-the-art overwrite, block erase and cryptographic erase methods for higher assurance, and media can be reused; Destroy uses physical destruction such as shredding, pulverizing and incinerating to render recovery infeasible, and media cannot be reused. The method must be chosen based on the sensitivity of the data, not the media type. File deletion, disk formatting and one-way encryption are not disposal methods — they leave most data intact and retrievable.

A legal hold (litigation hold) suspends normal deletion for data subject to litigation. The duty to preserve relevant evidence — paper or electronic — is triggered when civil litigation is commenced or reasonably anticipated, not only when a formal hold is issued; authorized access continues while deletion is prevented.

2.8 Design and implement auditability, traceability and accountability of data events

Accountability of data events requires defining event sources and the attributes each event must carry (identity, IP address, geolocation), then logging, storing and analyzing those events. A documented chain of custody — who handled evidence, when, and every transfer — preserves admissibility (developed further in Chapter 5).

2.9 Comprehend data protection of Artificial Intelligence (AI) and Machine Learning (ML) data

Training datasets and models are data assets subject to the same lifecycle and protections: dataset and model privacy (membership inference and model inversion can extract private training data from model outputs), and dataset and model security through validation and verification of data sources. Poisoned training data creates persistent backdoors — covered with AI threats in Chapter 4.

Key numbers & deadlines

Figure	Value	Source
CSA data security lifecycle phases	6	[4]
FIPS 140-3 supersedes 140-2 for new submissions	April 1, 2022	[3]
FIPS 140 security levels	4	[3]
Media sanitization actions (Clear, Purge, Destroy)	3	[5]

Key takeaways

• The CSA data lifecycle runs Create, Store, Use, Share, Archive, Destroy — and data does not always pass through every phase.
• A token has no mathematical path back to the original — only the vault can map it; ciphertext can be decrypted with the key. Protect the token vault.
• FIPS 140-3 is the current module-validation standard; Level 3 brings tamper-resistance and identity-based authentication, Level 4 brings tamper-active erasure.
• Hashing gives integrity and non-repudiation (via signatures), not confidentiality.
• Clear, Purge, Destroy escalate in assurance; choose by data sensitivity, and never rely on deletion or formatting as disposal.
• IRM protection travels with the data; legal holds trigger on anticipated litigation.

Chapter 2 quiz — 20 questions

Answer each question, then check the key that follows.

1. A security team models its data protections on the CSA Data Security Lifecycle. Which sequence of phases is correct?

• A. Create, Store, Share, Use, Archive, Destroy
• B. Create, Use, Store, Share, Destroy, Archive
• C. Create, Store, Use, Share, Archive, Destroy
• D. Store, Create, Use, Archive, Share, Destroy

2. During which phase transition does the CSA lifecycle note that storage typically occurs nearly simultaneously?

• A. Share to Archive
• B. Create to Store
• C. Use to Share
• D. Archive to Destroy

3. A retailer replaces stored credit-card numbers with randomly generated reference values that have no mathematical relationship to the originals. Which technique is this?

• A. Symmetric encryption
• B. Tokenization
• C. Hashing
• D. Data masking with format preservation

4. After a tokenization rollout, the security team hardens one particular database above all others. Which database, and why?

• A. The application database, because it holds the encrypted card numbers
• B. The logging database, because tokens appear in application logs
• C. The backup database, because backups contain the original plaintext
• D. The token vault, which holds the token-data mapping

5. A legacy billing application expects sixteen-digit card numbers in every record. Which token type integrates most easily with it?

• A. Non-format-preserving tokens, because they are shorter
• B. Encrypted tokens, because they can be decrypted by the application
• C. Hashed tokens, because they are deterministic
• D. Format-preserving tokens, because they keep the expected shape of the data

6. A federal agency is buying a new hardware security module in 2026 and requires a validated cryptographic module. Which standard applies to new submissions?

• A. FIPS 140-3
• B. FIPS 140-2
• C. FIPS 180-4
• D. FIPS 201-3

7. Which FIPS 140-3 security level is the minimum that requires production-grade equipment and externally tested algorithms?

• A. Security Level 4
• B. Security Level 2
• C. Security Level 1
• D. Security Level 3

8. Media sanitization is best defined as which of the following?

• A. Encrypting all data on the media with a strong algorithm
• B. Backing up the media before it leaves the facility
• C. Logging every access to the media in an asset inventory
• D. Making target data on the media infeasible to access

9. An administrator must decommission office laptops whose drives held internal documents, and the drives will be reused elsewhere in the company. Which sanitization action fits?

• A. Clear, using standard overwriting tools
• B. Degaussing, which is the only method approved for solid-state drives
• C. Shredding, which is required whenever drives leave a secure facility
• D. Reformatting, which removes the file allocation tables completely

10. A hospital retires servers that stored patient records and will redeploy the drives in a less sensitive system. Which action gives higher assurance than Clear while still allowing reuse?

• A. Purge, using laboratory overwrite or cryptographic erase
• B. Reformatting, which is faster and equally effective on modern drives
• C. Deleting, which removes directory entries so data is unrecoverable
• D. Defragmenting, which scatters file fragments beyond reconstruction

11. A data center must dispose of backup tapes containing highly confidential data that cannot be sanitized by overwriting. Which action is appropriate?

• A. Destroy, by shredding, pulverizing or incinerating
• B. Clear, followed by a full surface scan
• C. Purge, followed by redeployment to a low-security workload
• D. Degaussing, which allows the tapes to be safely resold

12. How should the sanitization method (Clear, Purge or Destroy) be chosen?

• A. By the age of the media, since older media holds data more tenaciously
• B. By the sensitivity of the data, not by the media type
• C. By the storage capacity, since larger drives need stronger methods
• D. By the manufacturer, since each vendor certifies one method

13. A technician is told to "dispose of" sensitive drives by deleting the files and formatting the disks. What is wrong with this plan?

• A. Nothing; deletion and formatting are approved sanitization methods
• B. Formatting is acceptable but deletion alone is not
• C. Deletion is acceptable but formatting alone is not
• D. Both leave most data intact and retrievable

14. Which pair of technologies does a CASB rely on to keep confidential data from leaving company-controlled systems?

• A. Firewalls and intrusion prevention systems
• B. Access control and data loss prevention (DLP)
• C. Sandboxing and URL filtering
• D. Tokenization and packet inspection

15. A contract document is emailed to an external law firm, yet the company can still prevent forwarding and printing. Which technology provides this persistent control?

• A. Transport Layer Security
• B. Information Rights Management
• C. Full-disk encryption
• D. Digital watermarking

16. Which IRM capability directly addresses a recipient forwarding a sensitive email to unauthorized parties?

• A. Automatic backup of the message to a journaling mailbox
• B. Spam filtering of the message before delivery
• C. Limiting who can copy, print or forward the content
• D. Archiving the message for eDiscovery

17. A system stores password verifiers so that any change to a stored value is detectable with very high probability, using a one-way function. Which mechanism is this?

• A. Symmetric encryption with a secret key
• B. Tokenization with a secure vault
• C. Cryptographic hashing producing a message digest
• D. Digital enveloping with a session key

18. A signed software update fails signature verification on a customer's machine. What does the failure prove, at minimum?

• A. The update was encrypted with the wrong key
• B. The update was sent to the wrong recipient
• C. The update's license has expired
• D. The message was altered after signing

19. A CASB's data-security pillar uses access control and DLP to achieve which outcome?

• A. Discovering undocumented cloud services in use across the company
• B. Preventing confidential data from leaving company-controlled systems
• C. Blocking malware downloads from sanctioned applications
• D. Automating incident response playbooks across tools

20. A company becomes aware of likely litigation over a contract dispute. When does its duty to preserve relevant emails and documents begin?

• A. When litigation is commenced or reasonably anticipated
• B. Only after a court issues a formal preservation order
• C. Only after opposing counsel sends a written request
• D. When the company's retention schedule next comes up for review

Answer key & explanations

1. C. The CSA Data Security Lifecycle runs Create, Store, Use, Share, Archive, Destroy, though data can bounce between phases and may skip stages. Putting Share before Use reverses the lifecycle's logic: data is stored, then used, and only shared afterward.^[4]

2. B. Storing is the act of committing digital data to a storage repository and typically occurs nearly simultaneously with creation. Archiving is a distinct later phase for long-term retention, not the near-simultaneous act of committing newly created data to a repository.^[4]

3. B. Data tokenization substitutes a sensitive data element with a non-sensitive placeholder called a token; tokens are generated randomly and lack any inherent or exploitable meaning or value. Hashing also breaks the mathematical link to the original, but hashes are deterministic — the same input always yields the same digest — while tokens are randomly generated placeholders.^[1]

4. D. A token vault securely stores the relationship between tokens and the original data, and protecting it is critical because it contains the key to retrieving sensitive information safely. Backups of the application database hold tokens, not the original data — only the vault's mapping can turn a token back into the sensitive value.^[1]

5. D. Format-preserving tokens keep the expected format of the data, so they integrate easily with systems that expect specific formats; non-format-preserving tokens do not resemble the original structure. Encrypted tokens can be decrypted, but decryption yields the original value, not a token shaped like a sixteen-digit card number — only format-preserving tokens keep the expected shape.^[1]

6. A. NIST's Cryptographic Module Validation Program stopped accepting FIPS 140-2 submissions for new validation certificates on April 1, 2022, so a module submitted for validation today is validated to FIPS 140-3, which supersedes FIPS 140-2. FIPS 140-2 is tempting because many deployed modules still carry 140-2 certificates, but no new 140-2 submissions are accepted and all FIPS 140-2 certificates are placed on the Historical List in September 2026.^{[2, 3]}

7. C. Level 1 requires production-grade equipment and externally tested algorithms — the baseline before higher levels add tamper-evidence, tamper-resistance and tamper-active protections. Level 2 adds tamper-evidence on top of the baseline — the production-grade-equipment and tested-algorithm requirement is already met at Level 1.^[3]

8. D. Media sanitization refers to a process that renders access to target data on the media infeasible for a given level of effort. Encryption protects data at rest or in transit but leaves it recoverable by anyone with the key; sanitization must render access to the target data infeasible.^[6]

9. A. Clear uses standard rewriting techniques and tools to provide moderate protection against simple, noninvasive data recovery techniques, and media can be reused after Clear sanitization. Reformatting only rewrites file-system structures, leaving most data intact and retrievable — it is not an approved sanitization method.^[5]

10. A. Purge uses state-of-the-art laboratory overwrite, block erase and cryptographic erase methods, provides a higher level of sanitization than Clear, and the media can still be reused afterward. Reformatting is neither approved nor equally effective: it leaves the data largely intact, which is exactly what higher-assurance sanitization must not do.^[5]

11. A. Destroy uses physical destruction techniques such as shredding, pulverizing and incinerating to render data recovery infeasible; it is used when media is beyond overwriting or holds highly confidential data, and media cannot be reused. Degaussing may sanitize magnetic tape, but it leaves the media unusable — the promise that degaussed tapes can be safely resold is false.^[5]

Sources cited in this excerpt

1. Tokenization vs. encryption. https://www.securitymetrics.com/blog/what-tokenization-and-how-can-i-use-it-pci-dss-compliance
2. NIST CSRC, FIPS 140-3 Transition Effort (Cryptographic Module Validation Program). https://csrc.nist.gov/projects/fips-140-3-transition-effort
3. FIPS 140-3 / CMVP cryptographic module validation. https://www.entrust.com/resources/learn/what-fips-140-3
4. CSA Data Security Lifecycle — six phases. https://github.com/cloudsecurityalliance/csa-guidance/blob/HEAD/Domain%205-%20Information%20Governance.md
5. Media Sanitization and Disposal Best Practices (U.S. Department of Education, Federal Student Aid). https://fsapartners.ed.gov/sites/default/files/2023-02/FSAMediaSanitization_BestPractices_508.pdf
6. NIST SP 800-88 Rev. 2, Guidelines for Media Sanitization. https://CSRC.NIST.GOV/pubs/sp/800/88/r2/final