Computer Engineering is coming soon

We're still writing and reviewing the notes for every subject, and we'll open it when every subject is ready. Meanwhile you can practise aptitude, ask Aria about any subject, or see the branches that are ready.

Risks in cloud computing

Risks in cloud computing: definitions, categories, quantitative measures (availability, risk score), mitigation controls and exam-focused worked examples for engineers.

Drafted with Aria, reviewed by the AiCanCode.org team · Updated . How we write and check our notes. Spotted an error? Use Give Feedback at the bottom of the page.

Why it matters

Cloud services host critical data and applications for organisations. An engineer uses these concepts to identify, measure and reduce the chances that data loss, service outages or compliance failures will harm users or business operations.

Key ideas

Start by defining the base terms. Define every term before use.

  • Cloud service: a remotely hosted computing resource (compute, storage, network, platform, or software) provided on demand over a network. I will refer to other topics by name (for example, "SaaS" or "virtualization") but not re-teach them here.

  • Risk: a statement of a possible adverse event described by two components: likelihood and impact. A formal definition: Risk = (Probability that an adverse event occurs) × (Consequence if it occurs). Assumption: likelihood and impact are on compatible scales (e.g., probability per year × monetary loss per event yields expected annual loss).

  • Threat: any potential cause of an unwanted incident that may harm a system or organisation. Example threats: DDoS, insider misuse, ransomware.

  • Vulnerability: a weakness in a system that can be exploited by a threat to cause harm (for example, an unpatched hypervisor vulnerability).

  • Impact: the measurable consequence of a successful attack or failure (monetary loss, data loss, reputation damage, safety). Define units when used (₹, USD, hours downtime).

  • Residual risk: the remaining risk after controls and mitigations are applied.

  • Attack surface: the set of reachable interfaces and components that an attacker can use to exploit a system (APIs, management portals, VM consoles).

  • Shared responsibility model: in cloud computing, a statement of which security tasks the cloud provider handles and which the customer must handle (provider may handle physical security; customer handles access controls, data encryption). Draw this as a two-column diagram labeled Provider / Customer and list responsibilities.

  • Multi-tenancy: multiple customers sharing the same physical infrastructure. Risk: noisy neighbours, data leakage.

  • Confidentiality, Integrity, Availability (CIA triad): standard security goals. Confidentiality means data is not disclosed to unauthorized parties. Integrity means data is accurate and not tampered with. Availability means services and data are accessible when needed.

  • SLA (Service Level Agreement): a contract between provider and customer stating service levels (availability %, response times) and penalties. An SLO (Service Level Objective) is a specific target for a defined SLI (Service Level Indicator); the SLA is the binding agreement that defines SLOs and associated remedies/penalties.

  • Durability: the probability that data remains intact and not lost over a period (often one year). Distinct from availability; e.g., object storage may offer 99.999999999% ("11 nines") durability while availability is lower.

  • MTBF (Mean Time Between Failures) and MTTR (Mean Time To Repair/Recover): reliability metrics used to compute availability. Definitions: MTBF = average time the system operates between failures; MTTR = average repair or recovery time to restore service after a failure. Note: MTTF (Mean Time To Failure) is used for non-repairable components; use MTTF in place of MTBF in those cases.

  • RTO (Recovery Time Objective) and RPO (Recovery Point Objective): disaster recovery targets. RTO = maximum acceptable time to restore service after outage. RPO = maximum acceptable data loss measured in time (how old backups may be).

  • Data locality: where data physically resides (which country/region). Legal/compliance risk arises when data crosses jurisdictions.

  • Vendor lock-in: difficulty moving data or applications from one cloud provider to another; a commercial and operational risk.

  • Supply-chain risk: risk introduced by third-party components (libraries, managed services) that the cloud stack depends on.

  • Identity and Access Management (IAM): processes and technologies for controlling who can do what in the cloud. Weak IAM is a top cloud risk.

  • Encryption & Key management: encryption protects confidentiality; key management decides who controls the cryptographic keys. If the provider manages keys, the customer must accept some trust; customer-managed keys reduce trust but increase operational complexity.

  • DDoS (Distributed Denial of Service): an attack aiming to make a service unavailable by overwhelming resources.

  • Insider threat: risk from authorised actors who misuse privileges.

Risk assessment process (common university-style steps):

  1. Asset identification (what to protect: data, VMs, containers, APIs).
  2. Threat and vulnerability identification.
  3. Likelihood estimation (qualitative: high/medium/low or quantitative: probability per year).
  4. Impact estimation (monetary, downtime hours, legal fines).
  5. Risk calculation (e.g. expected annual loss).
  6. Mitigation planning (controls chosen).
  7. Residual risk estimation and acceptance by stakeholders.

When an exam asks for a diagram, draw the following (labelled clearly):

  • A simple two-layer cloud diagram: top layer "Customer responsibilities" with boxes: IAM, data encryption, application configuration; bottom layer "Provider responsibilities" with boxes: physical security, hypervisor patches, network DDoS protection. Draw arrows showing shared points like APIs and management consoles. Title it "Shared responsibility model".

Formulas

Risk = Likelihood × Impact — Likelihood: probability (unit: 1, dimensionless) or frequency per year; Impact: monetary unit (₹, USD) or hours; Result: expected loss per period. Applies in quantitative risk analysis when probability and impact are known.

Availability = MTBF ÷ (MTBF + MTTR) — Availability is dimensionless (fraction) often expressed as a percentage. MTBF and MTTR have units of time (seconds, hours). Applies when failures are independent and repair restores full functionality. Equivalent downtime per year (hours) = (1 − Availability) × 8760 h; per 30-day month ≈ (1 − Availability) × 720 h.

Expected Annual Loss = Annual Probability × Monetary Impact — units: currency per year. Same assumptions as Risk formula. Also called Annualized Loss Expectancy (ALE).

Recovery Point Objective (RPO) = maximum tolerable data loss (time) — unit: time (s, h, days). Applies in backup strategy design.

Recovery Time Objective (RTO) = maximum tolerable service downtime (time) — unit: time. Applies in DR and failover planning.

Worked examples

Example 1 — Availability calculation (numerical) Problem: A cloud service has MTBF = 2,000 hours and MTTR = 4 hours. Compute availability. Solution steps:

  1. Formula: Availability = MTBF ÷ (MTBF + MTTR).
  2. Substitute numbers: Availability = 2000 ÷ (2000 + 4) = 2000 ÷ 2004.
  3. Compute: 2000 ÷ 2004 = 0.998003992... .
  4. As percentage: 0.998003992 × 100 = 99.8003992%. Answer: Availability ≈ 99.8004%. Comment: If SLA promises 99.9% availability, this system does not meet SLA (99.8004% < 99.9%).

Example 2 — Expected annual loss (quantitative risk) Problem: A particular cloud-stored dataset has a 2% annual probability of a breach; estimated impact on breach is ₹500,000. Compute expected annual loss. Solution steps:

  1. Use Expected Annual Loss = Probability × Impact.
  2. Substitute: 0.02 × 500,000 = 10,000. Answer: Expected annual loss = ₹10,000 per year. Interpretation: This helps decide investment in controls: spend up to ₹10,000/year to eliminate the risk, or less if mitigation only reduces probability.

Example 3 — Model answer to a typical long question (theory) Question: Describe the major risks of using a public IaaS cloud for sensitive enterprise data, and list mitigation measures. (Model answer) Model answer structure (full-marks includes both identification and mitigation):

  1. Identify risks: data confidentiality (multi-tenancy leakage, weak IAM), availability (DDoS, single region failures), compliance/data locality, vendor lock-in, insider threat, configuration errors (exposed S3 buckets). Define each in a sentence.
  2. For each risk give mitigations: encryption at rest and in transit with customer-managed keys (reduces confidentiality risk but adds key management burden); strong IAM with least privilege and MFA; network isolation (VPC, private endpoints); multi-region redundancy + async replication to meet RTO/RPO targets; regular backups with tested restores; DDoS protection service + autoscaling; logging and SIEM for detection; vulnerability management and timely patching; contractual SLA clauses, audits and right-to-audit for compliance.
  3. State residual risk and acceptance: after mitigations, estimate residual probability and impact and record in risk register; senior management accepts residual risk or demands further controls.
  4. Conclude: trade-offs (cost vs. risk) and reference the shared responsibility model. Mention that regular testing (penetration testing and DR drills) is required.

Common mistakes

  • Treating qualitative risk categories (high/medium/low) as numeric without defining scales.
  • Ignoring the shared responsibility model — students often assign provider responsibilities to the customer and vice versa.
  • Saying "encryption solves everything" without discussing key management and performance trade-offs.
  • Forgetting to state units for MTBF/MTTR, or mixing hours and seconds without conversion.

For your exam

Question types you will see: definitions (define risk, threat, vulnerability, SLA, MTTR), short notes (shared responsibility model, insider threat), derivations/calculations (compute availability from MTBF/MTTR, expected loss), diagrams (draw shared responsibility or attack surface), and long answers (risk assessment and mitigation plan). What a full-marks answer includes:

  • For definitions: concise definition and one short example.
  • For calculations: write the correct formula, substitute values with units, show arithmetic, and final answer with units and interpretation (compare with SLA or threshold).
  • For diagrams: clear labels and a brief explanation of each block/arrow.
  • For long answers: identify risks, quantify where possible, propose concrete mitigations, discuss residual risk and trade-offs, and mention testing and monitoring.

Quick check

  1. Define risk in one sentence.
  2. Give the formula to compute availability from MTBF and MTTR.
  3. What is the difference between RTO and RPO? Answers: Risk = probability × impact; Availability = MTBF ÷ (MTBF + MTTR); RTO is the maximum acceptable downtime, RPO is the maximum acceptable data loss measured in time.

(End of notes.)

Finished this topic? Mark it so your progress, study plan and readiness keep up.

Stuck on something here?