IAPP AIGP · Domain II.B · Exam Module

De-Identification & Privacy-Enhancing Technologies

Anonymisation, pseudonymisation, differential privacy and the whole PET toolbox — taught the way the exam tests it: by what survives, what falls away, and where the golden line sits.

1 interactive attack lab 15 flip cards 6-scenario matcher 8 exam traps GDPR · PDPA · EU AI Act
The golden rule, up front: pseudonymised ≠ anonymised. Pseudonymised data is still personal data. Only genuinely irreversible anonymisation exits the regime. Everything in this module hangs off that line.
Scroll to begin ↓
1
The de-identification spectrum

Where Does Your Data Sit?

De-identification is a spectrum, not a switch. Drag the slider and watch the legal status change as the link to the individual weakens. The golden line in the middle is the single most-tested boundary in this domain.

IDENTIFIEDDE-IDENTIFIEDPSEUDONYMISEDANONYMISED
Identified data
IN SCOPE — FULL REGIME
Direct identifiers present: name, NRIC, email, phone. Every data-protection obligation applies — lawful basis, purpose limitation, rights, breach notification.
Example: customer CRM records
The golden line (memorise this first)Anonymisation is the only position on the spectrum that exits the data-protection regime. Pseudonymisation is a safeguard, not an exit. Any exam answer claiming pseudonymised data is no longer personal data is wrong — in both the EU and Singapore contexts.[GDPR Art. 4(5); EDPB 01/2025]

Why "anonymised" is a claim, not a label

The EDPB's Article 29 Working Party Opinion 05/2014 treats anonymisation as a result to be tested against three criteria, not a box to tick: singling out (can one individual be isolated?), linkage (can two records about the same person be connected?), and inference (can an attribute be deduced with significant probability?). If any criterion is reasonably achievable, the data is not truly anonymised.[WP29 05/2014]

Singling out

Isolating one individual's record from the crowd — e.g., a row that is unique in the dataset. Countered by k-anonymity-style generalisation.

Linkage

Connecting two or more records that belong to the same individual across tables or datasets. The Weld case and the AOL scandal both died here.

Inference

Deducing an individual's attribute with significant probability from remaining data — the modern form is membership inference against ML models.

Exam anchorThe AIGP exam regularly asks which test establishes whether anonymisation has been achieved. Answer: the motivated intruder test — whether a reasonably competent outsider, with access to public resources, could re-identify individuals. This is the standard the EDPB applies to all three criteria above.[WP29 05/2014]
2
Interactive · Re-identification attack lab

Break the Anonymity — The Weld Attack

In 1997, Massachusetts Governor William Weld's medical records were re-identified from "anonymised" insurance data using just three quasi-identifiers: birth date, ZIP code and sex. This lab rebuilds that attack. Toggle the protections and watch the rows light up.

Run the linkage attack (join with voter register)
Hospital "anonymised" table
AgeDistrictSexDiagnosis
Voter register (public)
AgeDistrictSexName
34ClementiMWeld, W.
38ClementiMTan, K.
31West CoastFLim, S.
36JurongMNg, B.
33ClementiFOng, M.
39Bukit BatokMGoh, T.
Protected — no unique rows. With generalisation and suppression active, every row shares its quasi-identifier combination with at least 2 others. The attacker cannot single anyone out.
What just happenedWith protections off, the row 34 · Clementi · M is unique in the table — join it to the voter register and you have a name. That is the linkage attack (and singling out). Generalisation (age → band) and suppression (drop groups < 3) are the k-anonymity toolkit: every group now has ≥ 3 records. The dataset is 3-anonymous.[Sweeney 2002]

The k-anonymity family — the exam's favourite microdata models

k-anonymity (Sweeney, 2002): every combination of quasi-identifiers appears in at least k records. The classic failure: homogeneity attacks — if all k records share the same sensitive value (e.g., all 3 people in a group have the same disease), knowing someone is in the group reveals the disease.[Sweeney 2002]

l-diversity (Machanavajjhala et al., 2007): requires at least l distinct sensitive values per group — blocks homogeneity.[l-diversity 2007]

t-closeness (Li et al., 2007): the group's sensitive-value distribution must sit within distance t of the whole dataset's — blocks skewness attacks, which l-diversity misses when the global distribution is itself skewed.[t-closeness 2007]

ModelGuaranteeAttack it blocksLimitation
k-anonymityEach quasi-identifier group has ≥ k recordsRecord linkage / singling outHomogeneity attack — group can share one sensitive value
l-diversityEach group has ≥ l distinct sensitive valuesHomogeneity attackSkewness attack — skewed global distribution defeats "distinctness"
t-closenessGroup distribution ≈ global distribution (within t)Skewness attackUtility loss; hard to tune t; not a formal privacy guarantee
Exam memory hookk-anonymity = "k of us look the same". l-diversity = "l different secrets in the group". t-closeness = "the group's secret mix ≈ the whole dataset's mix". Each generation fixes the previous one's blind spot.[Sweeney 2002 → Li 2007]
▸ CASEThe AOL search scandal (2006) is the standard "de-identified but not anonymous" cautionary tale: AOL published 20 million "anonymised" search queries with user IDs. Reporters re-identified user 4417749 — Thelma Arnold of Lilburn, Georgia — purely from search content, including her name typed into her own searches. Query-level data is near-impossible to anonymise; the disclosure was pulled within days.Source: New York Times, "A Face Is Exposed for AOL Searcher No. 4417749" (Aug 2006)
3
Techniques that (aim to) remove the link

The Anonymisation Toolkit

Seven techniques, one goal: break the link to the individual. Each trades utility against residual risk — and none is a silver bullet. The exam wants you to recognise the technique from a description and know its weakness.

TechniqueWhat it doesTypical useKey weakness
AggregationCombines records into groups; reports only group statisticsOfficial statistics; research disclosuresSmall groups still leak individuals; loses record-level detail
Generalisation (banding)Broadens values: age → age band; postcode → districtk-anonymity datasets; public healthReduces precision; heavy generalisation destroys analytical value
Noise additionRandom perturbation so records distort but distributions surviveStatistical databases; location dataDeterministic attributes unaffected; can bias analysis
Data swappingExchanges attribute values between recordsCensus-style disclosure controlOriginal values reappear in combination; ordering attacks
Masking / redactionBlurs parts of values (NRIC S****123A)Screen redaction; test datasetsPartial identifiers often remain linkable — rarely anonymises alone
Hashing / keyed hashingReplaces identifiers with digests; keyed adds a secretJoining datasets without exposing raw IDsUnkeyed hashes of low-entropy values are trivially reversed
k-anonymity (model)Every record indistinguishable from ≥ k−1 others on quasi-identifiersPublished microdataFails on homogeneity and background-knowledge attacks
The hashing trap (classic exam question)"We hashed the NRICs, so the dataset is anonymous" → false. Unkeyed hashes of names/NRICs are trivially reversible by brute force over the small space of possible values. Keyed hashing (HMAC) resists brute force, but the key-holder can recompute the original — so the data is pseudonymised, not anonymised. Either way, obligations remain.[WP29 05/2014; EDPB 01/2025]

The Singapore framing

PDPC's Guide to Basic Anonymisation (updated 2024) walks organisations through a 5-step process: (1) identify direct and indirect identifiers, (2) apply de-identification techniques, (3) assess re-identification risk, (4) determine residual risk, (5) decide on safeguards. PDPC also publishes a free Data Anonymisation Tool. The Advisory Guidelines (revised 2024) clarify that obligations persist where re-identification remains reasonably possible — the label "anonymised" is never enough.[PDPC 2024]

Active recall · 15 high-yield terms

Flip Through the Terms

Click a card to flip. Each back face carries a definition plus an exam tip. Shuffle order in your head, not the deck — spaced repetition beats novelty tonight.

4
The safeguard that keeps obligations

Pseudonymisation — What It Is, What It Isn't

Pseudonymisation replaces direct identifiers with artificial identifiers so data can no longer be attributed to a data subject without the use of additional information — which must be kept separately and protected. That last clause is the whole exam: the key exists, so the link survives, so the data stays personal.[GDPR Art. 4(5); EDPB 01/2025]

TechniqueHow it worksReversibilityExam note
TokenisationIdentifier → random token in a secure mapping vaultReversible via the vaultMost common enterprise pattern; the vault is the risk point
EncryptionIdentifier encrypted; ciphertext is the pseudonymReversible with the keyDeterministic encryption allows matching across datasets — feature and risk
Keyed hash (HMAC)Identifier hashed with a secret keyRecomputable with the keyKey management is everything; unkeyed = broken
Random pseudonymsFully random replacement, no mathematical linkOnly via separately kept mappingDestroy the mapping → becomes anonymisation
The destroy-the-key escape hatchIf the separately held mapping or key is destroyed irreversibly, pseudonymised data can become anonymised — because the "additional information" that enables attribution no longer exists. This is the exam's favourite edge case: pseudonymisation + key destruction = anonymisation, but only if no other means of re-identification remain.[EDPB 01/2025; GDPR R26]

What EDPB Guidelines 01/2025 changed

The EDPB adopted Guidelines 01/2025 on Pseudonymisation on 16 January 2025. Key exam-relevant positions:[EDPB 01/2025]

Pseudonymisation ≠ anonymisation

Reiterated emphatically: pseudonymised data is personal data for any controller or processor holding the means of attribution, and remains in scope of GDPR (and by analogy PDPA).

It is a risk-reduction measure

Framed as a security and data-minimisation measure (Art. 32 and data-protection-by-design), not an exemption. It supports legitimate-interest balancing and compatible-purpose analysis.

Evidence over labels

Controllers must document the measures protecting the key, who has access, and the residual re-identification risk — the audit trail is what regulators will ask for.

Same-data pseudonymisation

Pseudonymising with the same key across datasets still permits linkage by the key-holder — a common mistake in research consortia and a documented exam scenario.

Pseudonymisation vs anonymisation — the master comparison

PseudonymisationAnonymisation
Reversible?Yes — via separately held key/mappingNo — link destroyed irreversibly
Still personal data?Yes — GDPR Art. 4(5); PDPA appliesNo, if genuinely achieved (high bar)
Role in lawSafeguard / security measure (Art. 32); data-minimisation supportExit from the regime; free reuse and sharing
Re-identification riskControlled by key security — but the key-holder can always linkResidual risk from linkage/inference; must be tested (motivated intruder)
Typical useAnalytics, research with follow-up, cross-dataset joins, test environmentsOpen-data publication, statistics, long-term archives
5
Privacy-Enhancing Technologies

The PET Toolbox — Full Comparison

PETs are the broader family: techniques that minimise personal data use while preserving functionality. From the mathematical (differential privacy) to the architectural (federated learning, secure enclaves). Expect scenario questions asking "which PET fits this constraint?"[AIGP BoK II.B; PDPC PET Guide 2023]

PETWhat it doesProtects againstKey tradeoffFlagship example
Differential privacy (DP)Adds calibrated noise so no single record materially changes outputs; governed by budget εMembership inference; singling outPrivacy–utility tradeoff; ε is a policy choiceUS Census Bureau 2020
Federated learning (FL)Training moves to the data; only model updates leave devicesCentralisation of raw dataGradients can leak — pair with DP + secure aggregationOn-device keyboard prediction (Gboard)
Synthetic dataGenerative models produce statistically similar artificial recordsRe-identification of source subjectsCan memorise outliers; inherits source biasNHS synthetic data pilots
Secure multi-party computation (SMPC)Parties compute jointly; each learns only the resultDisclosure of inputs during joint computationCommunication overhead; collusion analysisCross-bank fraud detection
Homomorphic encryption (HE)Computation on encrypted data; result decrypts correctlyDisclosure to the computing partyHigh computational overheadEncrypted cloud health analytics
Trusted execution environments (TEE)Hardware-isolated enclaves; data protected even from host OSHost compromise; insider accessHardware trust assumptions; side channelsConfidential computing (SGX/SEV/Nitro)
Zero-knowledge proofs (ZKP)Prove a statement ("age > 18") without revealing the valueOver-disclosure during verificationComputational cost; narrow use casesAge verification, credential attestation
Anonymisation suite§3 techniques: aggregation, generalisation, noise, k-anonymity familyDirect identification in published dataUtility loss; residual risk never zeroPDPC Data Anonymisation Tool
Pseudonymisation / tokenisationIdentifier replacement with separately held keys (§4)Casual exposure of identifiersKey-holder can always re-link; data stays in scopeNHS England pseudonymisation service

PET selection — the exam's decision logic

Q1 — Must data leave your control?

No → keep raw data local: federated learning or secure enclaves so only outputs/updates move.

Q2 — Joint computation needed?

Yes → SMPC (joint compute, no input sharing) or HE (compute on encrypted data by an untrusted party).

Q3 — Formal guarantee required?

Yes → differential privacy with documented ε. No → de-identification suite may suffice, but residual risk must still be assessed.

Q4 — Can analysis survive artificial data?

Yes → synthetic data for testing, augmentation, sharing. Verify fidelity per task — synthetic ≠ real for every use.

Q5 — Only one attribute needed?

Yes → zero-knowledge proofs for attribute verification without revealing the value.

Combination is normalPETs are complements, not competitors: federated learning + differential privacy + secure aggregation is the canonical stack for cross-institution ML. The exam rewards answers that combine PETs rather than picking a single winner.[McMahan 2017; AIGP BoK II.B]

DP vs FL vs synthetic data — the three most-confused PETs

Differential privacyFederated learningSynthetic data
Core mechanismCalibrated noise (formal ε guarantee)Distributed training; only updates travelGenerate artificial records from distributions
What it protectsAny single record's influence on resultsRaw data staying at sourceDirect subject mapping in released data
Primary weaknessUtility loss; ε is a value judgmentGradient leakage without DP/secure aggregationMemorisation of outliers; inherited bias
Best-fit scenarioStatistical releases over large datasetsModel training across devices/organisationsTesting, augmentation, dataset sharing
6
Easily confused — exam differentiators

Compare & Contrast

Five pairs the exam loves to blur. If you can fill these from memory, this domain is yours.

6.1 Anonymisation vs pseudonymisation (legal status)

AnonymisationPseudonymisation
Reversible?No — link destroyedYes — via separately held key
Still personal data?No, if genuinely achievedYes — GDPR Art. 4(5)
Legal effectExits the regimeSafeguard; risk reduction; remains regulated
How to proveMotivated intruder test / WP29 criteriaKey-security documentation and access control

6.2 Differential privacy vs federated learning

Differential privacyFederated learning
Protects againstIndividual-record influence on outputs (formal ε guarantee)Centralisation / movement of raw data
MechanismCalibrated noise in queries or trainingDistributed training; only updates shared
Key weaknessUtility loss; ε is a policy tradeoffGradient leakage; no formal guarantee alone
Flagship exampleUS Census 2020 disclosure avoidanceOn-device keyboard prediction (Gboard)

6.3 k-anonymity vs l-diversity vs t-closeness

k-anonymityl-diversityt-closeness
Requirement≥ k records per quasi-identifier group≥ l distinct sensitive values per groupGroup sensitive distribution within t of global
BlocksSingling out / record linkageHomogeneity attackSkewness attack
Fails againstHomogeneity; background knowledgeSkewness; background knowledgeBackground-knowledge extremes (partial)
CitationSweeney 2002Machanavajjhala 2007Li, Kifer, Machanavajjhala 2007

6.4 SMPC vs homomorphic encryption

SMPCHomomorphic encryption
ModelParties jointly compute; each holds a share of the resultOne party computes over encrypted data with the key held elsewhere
Data staysDistributed among parties (secret-shared)Encrypted at the computing party
Best forJoint analytics without sharing inputsOutsourced computation (cloud) on sensitive data
OverheadCommunication roundsComputation cost

6.5 Personal data vs special-category data (de-identification context)

Personal dataSpecial-category / sensitive data
GDPR regimeArt. 6 lawful basisArt. 9 conditions (plus Art. 6)
ExamplesName, contact, location, IDsHealth, race/ethnicity, biometrics, religion, sexual orientation
AI Act hookGeneral training-data governance (Art. 10(2))Bias detection exception with safeguards incl. pseudonymisation (Art. 10(5))
Exam significanceBaseline dutiesHigher bar; pseudonymisation named as an enabling safeguard
EU AI Act Art. 10(5) — the specific exam citationHigh-risk AI providers may process special-category personal data solely to ensure bias detection and correction, subject to strict necessity, appropriate safeguards, and technical measures — pseudonymisation is named explicitly as such a measure. This is the AI-Act hook for PETs and a favourite citation question.[EU AI Act (2024/1689), Art. 10(5)]
7
Interactive · Scenario matcher

Pick the Right Tool

Six real exam-style scenarios. Choose the best-fitting technique — instant feedback, score at the end. This is the closest thing to the actual scenario questions.

8
Self-test · Exam traps

Eight Exam Traps

TRUE / FALSE — instant feedback per question, score tallied at the bottom. Best score persists in this browser.

9
Exam-day readiness

Before Thursday — The Checklist

Tap each item as you master it. The goal: fill every box from memory, not recognition.

§
Citation ledger

Sources