Data Privacy and Data Security: Concepts, Worked Example and Exam Traps

Separate privacy decisions from security controls, then practise both ideas through a 12-record k-anonymity calculation and exam-style scenarios.

KnowledgeGate Team

Exam prep & CS education

Updated 31 Oct 20265 min read
Data Privacy and Data Security: Concepts, Worked Example and Exam Traps

A system can protect every file with encryption and still violate privacy by using student data for an undisclosed purpose. The reverse also happens: a clear privacy promise means little when the storage bucket is publicly readable. The difficulty is that data privacy and data security overlap without meaning the same thing.

Data privacy governs collection purpose, necessary fields, permitted use and sharing, retention, and the data subject's control. Data security protects data against unauthorised disclosure, alteration, destruction, or loss of access.

An encrypted student database used for undisclosed advertising may be secure against outsiders but not privacy-respecting. A portal with a careful privacy notice but a publicly readable storage bucket has a security failure.

Privacy sets the permitted purpose and boundary. Security supplies controls that enforce it. They overlap without becoming interchangeable. For Indian legal duties, check the Digital Personal Data Protection Act, 2023 on India Code; purpose, consent and retention duties depend on the law, not on a security-control checklist.

Classify the data before choosing a control

Start by identifying what each value can reveal:

  • student17@example.com is a direct identifier because it directly points to an account or person.

  • Age 23 and PIN code 110002 are quasi-identifiers. Either may look harmless alone, but a combination can narrow the possible person.

  • Screen reader is a sensitive support need.

  • The 12-record support table is the dataset being protected.

Data is in transit while a CSV moves to an authorised reviewer, at rest while stored, and in use while the reviewer filters it. Encryption can protect transfer or storage, but an authorised process normally needs plaintext during use.

Collect only necessary fields, separate direct identifiers, grant least-privilege access, encrypt appropriate copies and transfers, log access, keep recoverable backups, and delete or aggregate data when its purpose ends.

Concept map contrasting privacy decisions and security controls, with age, PIN, support need, 12-row access, encryption, logs, integrity checks, backups, and authorised use only.

Worked example: reduce re-identification risk in a 12-record release

Suppose a support-planning team removes the email column before releasing this hypothetical table. Age and PIN remain quasi-identifiers.

Row

Age

PIN

Support need

R1

22

110001

None

R2

22

110001

Extra time

R3

23

110002

None

R4

23

110002

Screen reader

R5

24

110003

Extra time

R6

24

110003

None

R7

25

110004

Screen reader

R8

25

110004

None

R9

26

110011

Extra time

R10

26

110011

None

R11

27

110012

Screen reader

R12

27

110012

None

Group records by exact (Age, PIN). The pairs (22, 110001), (23, 110002), (24, 110003), (25, 110004), (26, 110011), and (27, 110012) each appear twice.

The smallest equivalence class has 2 records:

k = min(2, 2, 2, 2, 2, 2) = 2

Under a simplified equal-likelihood assumption, the worst-case record-linkage probability is:

1/k = 1/2 = 0.50 = 50%

Generalise ages 22-25 and 26-27 into two bands. Replace PINs 110001-110004 with 11000* and 110011-110012 with 11001*. The new classes contain R1 to R8 and R9 to R12, giving sizes 8 and 4.

k = min(8, 4) = 4

The simplified worst-case linkage probability becomes:

1/k = 1/4 = 0.25 = 25%

This costs utility: six exact ages become two bands and six exact PIN codes become two prefixes. K-anonymity reduces privacy risk. It is not encryption or a guarantee against every inference attack.

Before-and-after table showing six size-2 Age and PIN groups reduced to two bands and prefixes, producing group sizes 8 and 4 with k increasing from 2 to 4.

Map threats to confidentiality, integrity, availability and privacy

CIA classifies security objectives. Privacy also asks whether an authorised use stays within its permitted purpose.

Case

Property at risk

Matching control

An attacker reads all 12 support rows

Confidentiality, with possible privacy harm

Access control and encryption

A clerk changes R9's score from 72 to 27

Integrity

Authorisation, audit trails and integrity checks

An outage blocks the portal for 90 minutes

Availability

Redundancy and tested recovery

An authorised team reuses 12 email addresses for ads

Privacy purpose

Purpose limitation and data minimisation

Network protections are one security layer. They do not decide whether collection or later use is permitted.

Common traps that turn correct definitions into wrong answers

Replacing student17@example.com with T017 is pseudonymisation if a separate lookup restores identity. The token remains personal data in this example. Deleting the lookup does not prove anonymity if age and PIN can still single out someone.

Keep these pairs separate:

  • Authentication asks who the user is. Authorisation asks what that user may do. Accountability records what was done.

  • Encryption is reversible with the right key. A cryptographic hash is designed as a one-way digest.

  • A backup supports recovery and availability, but does not by itself prevent disclosure.

First name the asset and permitted purpose. Identify the threat, classify the property at risk, then choose the control. Avoid absolutes such as "encrypted means private" or "consent makes every later use valid".

How exams turn the topic into questions

Expect definition matching, privacy-versus-security scenarios, CIA-to-control mapping, and small equivalence-class calculations such as k = min(8, 4) = 4. No fixed frequency is implied for any named exam.

For a 20-second scenario method:

  1. Underline the action: read, changed, unavailable, or reused.

  2. Identify whether the actor was authorised.

  3. Check the stated purpose.

  4. Select confidentiality, integrity, availability, privacy, or a combination.

Self-check: why is authorised advertising reuse a privacy problem rather than a confidentiality breach? Because the team had access, but the new purpose crossed the permitted-use boundary. No outsider needed to read the data.

Use the UGC NET Computer Science syllabus areas to place this subtopic in the wider Paper 2 map, then practise with the UGC NET complete CS test series.

The short version and what to study next

  • Privacy governs permitted collection and use.

  • Security protects data.

  • Confidentiality, integrity and availability classify security objectives.

  • Quasi-identifiers can re-identify records after names are removed.

  • One control never solves every risk.

Now change the smaller generalised class from 4 records to 3. The result is k = min(8, 3) = 3, and the simplified worst-case linkage probability is 1/3 = 0.333..., or about 33.3%. This variant shows why shrinking the smallest class increases linkage risk.

For topic prioritisation, read UGC NET Computer Science high-yield topics. Continue with the structured NTA UGC NET Paper 2 course, or explore the broader UGC NET category before choosing the next area to revise.