Privacy enhancing security COM3030

What is Privacy as Control?

User participation, Transparency, Accountability and Organisational compliance

What is Privacy as Transparency?

Improve user agency, Aid decision making and transparency of social impact

What is Privacy as confidentiality

Minimize data disclosure, distribute trust

Who is an adversary?

The “Others“, any person or entity outside of your system

Why is Surveillance not inherently Security?

Surveillance may not be effective for smart adversaries, Surveillance tools can be abused and can be subverted for crime/terrorism

What makes Privacy a Security property?

For individuals its about protection against crime and control over one’s information. For companies its about protection of trade secrets and internal operations. For governments its about protection of national secrets and law enforcement investigations.

What are the privacy properties?

Confidentiality, Pseudonymity, Anonymity, Unlinkability, Unobservability, Plausible deniability

What is the No Free Lunch Theorem?

The No Free Lunch Theorems state that any one algorithm is optimal for one solution cannot be universally superior to any other algorithm.

What is Pseudonymity?

A user may use a resource or service without disclosing its identity, but can still be accountable for that use.

What is Anonymity?

Anonymity is the state of being unidentifiable within a set of all possible subjects that might cause an action. Decoupling identity and action.

Anonymity vs Privacy

Privacy is the control over personal information and the ability to keep it confidential, whereas anonymity is being unknown and unidentifiable.

What is Unlinkability?

The ability to make multiple uses of resources or services without others being able to link these uses together. Decoupling two actions from one user.

What is Unobservability?

Using a resource or service without others, especially third parties, being able to observe its use. Decoupling observation from action. Implies anonymity

What is Plausible Deniability?

Making it impossible to prove a user knows, has done or has said something. Decoupling observation from true action

What is a One Time Pad?

A single use, random string used for encryption the same length as the string

What is Computational Security?

Computational security is the property of a cryptographic system where breaking the encryption or gaining meaningful information is theoretically possible but practically infeasible due to resource constraints, such as time, computational power, or money

What is Public key encryption?

An asymmetric cryptography method that uses a mathematically linked public and private key to secure data and enable confidential communication.

What is Encryption’s requirements, pros and cons?

Encryption requires a secret key. It has a strong guarantee at the cost of efficiency and flexibility. Information leakage is negligible.

What is Obfuscation?

Obfuscation has no secret. It limits information leakage by adding and removing obfuscation noise.

What is the difference between a key attribute, sensitive attribute and a quasi-identifier?

A key attribute is a unique identifier, a sensitive attribute is one that is legally or ethically protected, and a quasi-identifier is an attribute that does not uniquely identify a person on its own but can reveal identity when combined with other data.

What are the Data Anonymisation properties?

k-anonymity, t-closeness, l-diversity among others.

What are the four data anonymisation techniques?

Data Generalization, Data Suppression, Dummy data addition and Data Perturbation

What is Data Generalisation?

Coarsening an attribute to a more general value through reduction or bucketisation. E.g. changing a ZIP code from 51300 to 51*** or an age of 21 to 10-30.

What is Data Suppression?

Suppressing information by not releasing values e.g Hiding 50% of values in a column for count suppression, or 100% for full suppression.

What is the difference between data generalisation and suppression?

Generalisation is replacing values with a less precise one, whereas suppression is not releasing the values.

What is Dummy Addition?

Adding fake data to what the adversary receives. This leads to errors in adversary’s inference and must be indistinguishable from real data.

What is the difference between data suppression and data anonymisation?

Data Suppression is removing values whereas data anonymisation is altering values.

What is Data Anonymisation?

Decoupling identity from information using quasi-identifiers and k-anonymity.

What is k-anonymity?

Modifying attributes such that any possible grouping has at least k-values per quasi-identifier

What is l-diversity?

Every grouping has at least l sensitive values

What is T-closeness?

How close is the distribution of sensitive values in a particular table from distribution over all sensitive values (from public sources).

What is Data Perturbation?

Adding noise, shuffling values or altering numerical values within certain limits and preserving statistical properties.

What is differential privacy?

When an adversary can make the same inference about any individual private information, whether or not that individual’s private information is included in the input of the analysis.

What is a Shamir Threshold Secret Sharing?

A secret sharing scheme splits a secret into shares under some policy, such that the threshold shares can later be used to recover the secret. At least t+1 participants are needed.

What is a weighted structure for secret sharing? Ψ 𝐴 = 𝑡 < 𝑤𝑒𝑖𝑔ℎ𝑡 𝐴 𝑎𝑛𝑑 𝐴 ⊆ 𝑈

The sum of all weights assigned to set 𝑥 is at least 𝑡 + 1. Each participant has a positive weight based on their role.

What is Hierarchical structure for Secret Sharing?

Participants are divided into disjointed levels according to their importance, with the lowest being 1.

How can a dishonest Shamir dealer cheat?

By using a polynomial with the wrong degree or distributing inconsistent shares, causing unauthorised recovery or inconsistent secrets.

How can a dishonest participant cheat during recovery?

By submitting a fake share, preventing honest participants from recovering the original secret.

What is secure multi-party computation?

A protocol allowing parties to compute a public function over private inputs without revealing those inputs.

What is oblivious transfer?

A two-party primitive where the receiver obtains selected data without revealing the choice, and the sender's unselected data remains hidden.

What are OT correctness and privacy?

Correctness means intended outputs are obtained; privacy means each party learns nothing beyond its intended output.

What is receiver security in OT?

A sender cannot determine which input the receiver selected.

What is sender security in OT?

A receiver cannot learn the sender's unselected input.

Name five outsourcing techniques from the module.

Oblivious transfer, private information retrieval, privacy-preserving machine learning, searchable encryption, and secret sharing.

What is perfect privacy in secret sharing?

Unauthorised shares reveal no information about the secret; the adversary's advantage is zero.

What is computational privacy for secret sharing scheme?

The adversary's advantage is negligible, so it can guess the secret bit with only approximately 50% probability.

What is a key advantage of zero-knowledge age verification?

A user can prove they meet an age requirement without revealing unnecessary identity information.