Jahziah.
العربية

STUDY GUIDE · GKU 8

Ethics, privacy and security study guide (Artificial Intelligence)

An AI system is software that makes or shapes decisions about people from data about people, so it inherits every ethical and security problem of ordinary software and adds its own. The first half of this unit is about the people affected: what ethical and privacy issues AI raises and how Saudi Arabia's SDAIA AI Ethics Principles and Personal Data Protection Law frame them; how to protect privacy by design, with anonymisation, k-anonymity and differential privacy; and how to measure and reduce unfairness, with group metrics you can compute from a table, and how explainability supports trust. The second half is security: confidentiality, integrity and availability; risks, threats, vulnerabilities and attack vectors; authentication, authorisation and access control; what makes an AI system trustworthy; and how attacks that are specific to machine learning — poisoning, adversarial examples, model extraction — are prevented. Exam questions here are mostly scenarios: name the property, the threat or the control.

Continue in the app

Structured by the official ETEC subject standard for this topic: every listed topic and learning outcome, with formulas, figures and worked examples. The full sections are available to members in the app.

SKU 8.1 · Ethical issues and privacy

  1. Ethical and privacy issues of AI, and the Saudi frame

    The recurring ethical concerns of computer-based and AI systems — bias, privacy, surveillance, opacity, accountability, safety, manipulation and effects on work — and the principles that answer them, including SDAIA's AI Ethics Principles and the Personal Data Protection Law.

  2. Designing privacy protection: anonymisation, k-anonymity and differential privacy

    Why removing names does not anonymise data, how k-anonymity and its weaknesses are checked on a table, how differential privacy adds calibrated noise with a provable guarantee, and the design methods — minimisation, on-device and federated learning — that keep data from moving at all.

  3. Fairness, bias, trust and explainability

    Where bias in AI comes from, the group-fairness metrics computed from a table (demographic parity, disparate impact, equal opportunity), why deleting the protected attribute does not remove bias, how reweighing and threshold adjustment mitigate it, and how explanations build justified trust.

SKU 8.2 · Security

  1. Confidentiality, integrity and availability in AI systems

    The three core security properties, what each means for the data, the model and the service of an AI system, the controls that protect each, and how to name the property an incident violates.

  2. Risk, threats, vulnerabilities and attack vectors

    The vocabulary of security risk — asset, threat, vulnerability, attack vector, risk, control — applied to AI systems, with qualitative risk scoring on a likelihood–impact matrix and quantitative annualised loss expectancy to decide whether a control is worth its cost.

  3. Authentication, authorisation, access control and trustworthy AI

    How a system decides who you are and what you may do, the main access-control models, how credentials leak in AI applications, and what makes an AI system trustworthy rather than merely trusted.

  4. Attack prevention for machine-learning systems

    The attacks specific to ML — data poisoning, adversarial (evasion) examples, model extraction, model inversion and membership inference — alongside classic application attacks, why a linear model is easy to fool in high dimensions, and the matching defences.

MORE ARTIFICIAL INTELLIGENCE GUIDES

All study guides

Read the full guide beside your practice questions.

Open the app