Detective Tooling Bundle

Amazon CloudWatch and AWS Trusted Advisor — the generic, foundational observability and best-practice-checking layer that almost every other security service in this study set quietly relies on or gets compared against. Neither is a dedicated "security service" the way GuardDuty or Inspector are, but both show up constantly as integration points, distractors, and standalone exam targets in their own right.

Note: IAM Access Analyzer — originally grouped under "Detective Tooling" — is covered in the IAM & Governance Bundle, since it's structurally an IAM capability. This file focuses on CloudWatch and Trusted Advisor.

Detective — primary role Governance — best-practice checks Responsive — alarm-driven automation
ML-powered
CloudWatch Logs anomaly detection
500
Log anomaly detector quota per account
482
Trusted Advisor checks at Business Support+
56
Checks available on every support plan

The Core Mental Model — Generic Observability vs Specialized Detection

Two Tools, One Shared Role: The Foundation Everything Else Builds On
⚠️ The Recurring Exam Theme

Questions here test whether you know these are GENERIC tools layered underneath the specialized security services already covered, not replacements for them. CloudWatch alarms can detect a metric threshold breach, but they don't understand "this is a credential compromise" the way GuardDuty does — you have to build that logic yourself via metric filters/alarms. Trusted Advisor's security checks are real but limited and support-plan-gated — it complements, but doesn't replace, Config or Security Hub CSPM for deep, customizable compliance evaluation.

Exam Domain Mapping

DomainWhere This Bundle Shows Up
Security Logging & MonitoringThe centerpiece — metric filters, alarms, Logs Insights, anomaly detection
Threat Detection & Incident ResponseCustom alarm-driven response patterns, EventBridge integration for automated remediation
Management & Security GovernanceTrusted Advisor's security category checks, Organizational View for multi-account visibility

Decision Tree — Mental Model

Threat

A metric crossing an unsafe threshold; unusual patterns in log volume/content that no specialized service is watching for; basic AWS misconfigurations (public S3 buckets, open security group ports, missing root MFA) going unnoticed

Security Goal

Generic, broad-coverage observability and best-practice checking underneath/alongside the specialized detection services

AWS Service

The Detective Tooling Bundle

CloudWatch metrics + alarms CloudWatch Logs Insights + anomaly detection Trusted Advisor security checks Organizational View
Implementation

Build metric filters/alarms on security-relevant log patterns; enable CloudWatch Logs anomaly detection on critical log groups; ensure Business Support+ (or higher) for full Trusted Advisor security check coverage; enable Organizational View for multi-account rollup.

Monitoring

CloudWatch alarms → SNS/EventBridge; Trusted Advisor refreshed checks (automatic on Business Support+, manual on Basic/Developer) surfaced via API/console/EventBridge.

Remediation

EventBridge-triggered Lambda on alarm/check state change; manual remediation guided by Trusted Advisor's specific recommendation per check.

Final Summary

Must Memorize
  • CloudWatch is the generic substrate other security services route into — it has no inherent security-specific intelligence of its own beyond what you build (alarms/filters) or its own ML anomaly detection features
  • CloudWatch Logs anomaly detection is ML-powered, free to create detectors, with a default quota of 500 per account
  • Trusted Advisor's FULL check set (482 total) requires Business Support+ or higher; free/Basic-Developer tiers get only 56 checks (service limits + select security)
  • Trusted Advisor security checks (free tier) include: public S3 buckets, public EBS/RDS snapshots, unrestricted security group ports, root MFA status
  • Basic/Developer Support plans require MANUAL refresh of Trusted Advisor checks; Business Support+ and above refresh automatically
Must Understand
  • Why CloudWatch alarms require YOU to define the threshold/pattern, unlike GuardDuty's pre-built ML threat detection
  • The conceptual overlap and distinction between Trusted Advisor, Config, and Security Hub CSPM — all evaluate against best practices/rules, at different depths of customization
  • The Organizational View pattern for rolling up Trusted Advisor findings across an entire AWS Organization
Can De-prioritize
  • Exact dollar pricing per alarm type
  • Full list of all 482 Trusted Advisor checks
  • Console UI navigation specifics

Exam appearance probability: MEDIUM

Core Services — Deep Dive

CloudWatch Core Concepts The generic substrate
MetricsTime-series data points published by AWS services or custom applications — the raw measurements alarms evaluate
Metric filtersPattern-matching rules applied to CloudWatch Logs that turn matching log lines into a CloudWatch METRIC (e.g. count of "AccessDenied" occurrences per minute)
AlarmsWatch a metric (standard threshold OR anomaly detection band) and transition between OK/ALARM/INSUFFICIENT_DATA states, triggering notifications/actions on state change
Composite alarmsCombine multiple existing alarms using AND/OR logic into one higher-level alarm — reducing noise from correlated, simultaneous alarm firing
  • The classic security pattern: trail delivers events to CloudWatch Logs → a metric filter matches a sensitive pattern (root account usage, unauthorized API calls, security group changes) → an alarm fires → SNS notifies the team or EventBridge triggers automated response.
CloudWatch Metric Anomaly Detection ML-based banded thresholds
What it doesBuilds a statistical model of a metric's normal behavior over time, generating an expected "band" of normal values — alarms can trigger when the metric falls OUTSIDE that band, rather than against a single fixed threshold
Anomaly detection thresholdA configurable sensitivity value — higher creates a thicker, more tolerant "normal" band; lower creates a thinner band that alarms on smaller deviations
Metric math supportAnomaly detection can also be applied to METRIC MATH expressions (combinations/calculations across multiple metrics), not just raw single metrics — including support for Prometheus-compatible metrics via PromQL functions
Model warm-upFor a brand-new model, it can take up to 3 hours before the actual anomaly detection band appears/becomes usable
  • "Detect an unusual deviation in a custom application metric, without manually setting and tuning a fixed numeric threshold" → CloudWatch metric anomaly detection.
CloudWatch Logs, Insights & Logs Anomaly Detection High — ML-powered, exam-fresh refinements
CloudWatch Logs InsightsA purpose-built query language for ad-hoc, interactive analysis directly against log data
Log classesStandard ($0.50/GB) vs Infrequent Access ($0.25/GB, half the cost) — as of early 2026, Infrequent Access logs support FULL Logs Insights querying, making it the recommended default for application logs that don't need real-time alarming
Log anomaly detectionML-powered — automatically infers PATTERNS in log data (treating dynamic tokens like error codes, IPs, timestamps, request IDs as variable parts of a pattern), trains a model on those patterns, then flags significant fluctuations as anomalies after the training period; each anomaly is assigned a priority
CostCreating log anomaly detectors does NOT incur a charge
Default quotaIncreased from 10 to 500 log anomaly detectors per account (Oct 2024)
On-demand anomaly queriesThe anomaly command in Logs Insights queries lets you find unusual patterns ON-DEMAND (not just via a continuously-running detector) — detecting five types of anomalies including pattern frequency changes, new patterns, and token variations
EncryptionLog anomaly detector models and their results can be encrypted with AWS KMS
  • "Identify a brand-new, never-before-seen error pattern appearing in our application logs, without manually writing a filter pattern for something we don't yet know to look for" → CloudWatch Logs anomaly detection (it specifically catches NEW patterns, not just frequency changes in known ones).
AWS Trusted Advisor Support-plan-gated best-practice checks
CategoriesCost Optimization, Performance, Security, Fault Tolerance, Service Limits, Operational Excellence
Check status colorsGreen (no problem), Yellow (investigation recommended), Red (action recommended), Gray (excluded items)
Free / Basic / Developer Support tierAll accounts get 56 checks: full Service Limits category, plus SELECT Security checks (public S3 buckets, public EBS/RDS snapshots, unrestricted security group ports, root account MFA status)
Business Support+ and aboveUnlocks the FULL 482-check set across all six categories
Refresh behaviorBusiness Support+ and above: automatic refresh. Basic/Developer Support: MUST manually refresh Security category checks
Config-powered checks64+ checks (added 2023) are powered by underlying AWS Config managed rules — automatically reflecting findings if Config is enabled, available to Business/Enterprise-tier customers
Organizational ViewAggregates Trusted Advisor findings across an ENTIRE AWS Organization — requires Organizations with all features enabled plus a supported Support plan
  • Support plan restructuring (2026): Developer Support, Business Support, and Enterprise On-Ramp are being discontinued January 1, 2027, with customers steered toward Business Support+ (delivering AI-powered assistance) or Enterprise Support (with a reduced $5,000 minimum, down from $15,000). This is a live transition worth knowing exists, even if exact pricing isn't exam-critical.
  • "Determine, at no incremental support-plan cost, whether our root account has MFA enabled and whether any S3 buckets are public" → Trusted Advisor's free-tier security checks satisfy this without needing Business Support+.

AWS Exam Thinking

Requirement → Keywords → Expected Answer → why every distractor fails.

Alert the team in near real-time when "root account login" appears in CloudTrail logs
root account usagenear real-time alert
Expected Answer

CloudTrail → CloudWatch Logs → metric filter → alarm → SNS

DistractorWhy it's wrong
GuardDutyCould ALSO detect anomalous root usage via its own ML, but the SPECIFIC, deterministic "this exact pattern occurred" alerting is the textbook metric-filter-and-alarm pattern
Trusted AdvisorChecks root MFA STATUS (a static configuration check), not real-time LOGIN EVENT alerting
Catch a brand-new, previously unseen error pattern in application logs without writing a filter for it in advance
never-before-seen patternno pre-written filter
Expected Answer

CloudWatch Logs anomaly detection

DistractorWhy it's wrong
A manually written metric filterBy definition requires you to already know the pattern to write the filter — can't catch something genuinely new
Check, at zero incremental support-plan cost, whether any S3 buckets are publicly accessible or root MFA is missing
no incremental costbasic security posture check
Expected Answer

Trusted Advisor (free-tier security checks)

DistractorWhy it's wrong
Security Hub CSPMA real, deeper option, but introduces its own per-check evaluation cost — not the "zero incremental cost" answer the requirement specifically calls for
Get a single, consolidated view of Trusted Advisor findings across every account in an AWS Organization
org-wide rollupconsolidated view
Expected Answer

Trusted Advisor Organizational View

DistractorWhy it's wrong
Manually checking Trusted Advisor in each account individuallyDoesn't scale and defeats the purpose of a consolidated view that Organizational View provides natively

Integrations

AWS CloudTrail
WhatA trail can stream events to CloudWatch Logs, enabling metric filters/alarms on specific API activity patterns
Amazon SNS
WhatThe standard notification target when a CloudWatch alarm transitions to ALARM state
Amazon EventBridge
WhatBoth CloudWatch alarm state changes and Trusted Advisor check status changes can be routed through EventBridge for automated, programmatic remediation workflows
AWS KMS
WhatEncrypts CloudWatch Logs anomaly detector models and their results
AWS Config
WhatPowers 64+ Trusted Advisor checks directly — Trusted Advisor reflects findings from deployed Config managed rules automatically when Config is enabled
AWS Organizations
WhatRequired (with all features enabled) for Trusted Advisor's Organizational View, consolidating findings across every member account

Costs, Limits & Quotas

Pricing Model

Standard resolution alarms~$0.10/alarm/month (60-second evaluation)
High-resolution alarms~$0.30/alarm/month (10 or 30-second evaluation)
Anomaly detection alarms~$3.00/alarm/month
Composite alarms~$0.50/composite alarm/month
CloudWatch Logs — Standard class~$0.50/GB after a 5GB/month free tier
CloudWatch Logs — Infrequent Access class~$0.25/GB — half of Standard, now with full Logs Insights query support (early 2026+)
CloudWatch Logs anomaly detectorsNo charge to create
Trusted AdvisorIncluded with the relevant AWS Support plan — no separate per-check charge

Common Cost Mistakes

Cost Optimization

Limits & Quotas

Log anomaly detectorsDefault quota of 500 per account (increased from 10 in October 2024)
New anomaly model warm-upUp to 3 hours before the detection band appears for a brand-new metric model
Trusted Advisor free-tier checks56 checks total (Service Limits + select Security) on Basic/Developer Support
Trusted Advisor full check set482 checks, requiring Business Support+ or higher
⚠️ Exam trap

On Basic or Developer Support, Trusted Advisor's Security category checks do NOT refresh automatically — they must be manually refreshed, meaning a stale check result could mask a recently introduced misconfiguration if no one remembers to refresh it.

Best Practices & Common Exam Traps

8 — Best Practices

Must Know
  • CloudWatch has no inherent security intelligence — alarms/filters require YOU to define what to watch for
  • CloudWatch Logs anomaly detection is ML-powered, free to create, catches genuinely NEW patterns, not just threshold breaches
  • Trusted Advisor's full 482-check set requires Business Support+; free tier gets only 56 (Service Limits + select Security)
  • Basic/Developer Support requires MANUAL refresh of Trusted Advisor Security checks
  • Organizational View requires Organizations with all features enabled + a supported Support plan
Good Practice
  • Build metric filters/alarms on CloudTrail-derived security-relevant log patterns (root usage, unauthorized API calls)
  • Enable CloudWatch Logs anomaly detection on critical, high-signal log groups
  • Route both CloudWatch alarms and Trusted Advisor check changes through EventBridge for automated response
  • Verify Support plan tier covers the Trusted Advisor security checks your compliance posture actually depends on
Advanced Practice
  • Use composite alarms to reduce noise from correlated individual alarm firings
  • Use the on-demand anomaly Logs Insights command for ad-hoc investigation during incident response, not just continuous detectors
  • Leverage Config-powered Trusted Advisor checks for automatic reflection of Config managed rule findings without separate setup
  • Use Infrequent Access log class with full Logs Insights support for cost-efficient long-term log retention without sacrificing query capability

9 — Common Exam Traps

MisconceptionReality
"CloudWatch automatically knows what 'malicious' activity looks like, like GuardDuty does"False — CloudWatch alarms/filters only watch for whatever pattern/threshold YOU explicitly define; its anomaly detection is statistical, not threat-intelligence-driven
"All Trusted Advisor checks are available on every Support plan"False — only 56 of 482 checks are available below Business Support+; the rest require that tier or higher
"Trusted Advisor checks always refresh automatically"Only true on Business Support+ and above — Basic/Developer Support requires manual refresh of Security category checks
"Trusted Advisor is a full replacement for AWS Config or Security Hub CSPM"It's broader (covers cost/performance too) but shallower on security-specific depth and customization — complementary, not a replacement
"CloudWatch Logs anomaly detection costs extra to set up"Creating log anomaly detectors does not incur a charge — standard CloudWatch Logs ingestion/storage costs still apply, but the detector itself is free

Lookalike Comparisons

ComparisonWhat actually differs
CloudWatch vs GuardDutyGeneric, define-it-yourself metric/log monitoring vs purpose-built, pre-trained ML threat detection across curated AWS-native data sources
CloudWatch metric anomaly detection vs Logs anomaly detectionStatistical "normal band" around a numeric METRIC vs ML-inferred PATTERN analysis of log TEXT, including detecting genuinely new patterns
Trusted Advisor vs AWS ConfigBroad, opinionated, support-plan-gated best-practice checks across 6 categories (cost/performance/security/etc.) vs a deep, fully customizable configuration-rule evaluation engine you can extend with your own custom rules
Trusted Advisor vs Security Hub CSPMGeneral AWS best practices, broader scope beyond just security vs standards-mapped (CIS/PCI/NIST/FSBP), security-specific, continuously-scored compliance framework
CloudWatch Logs Standard vs Infrequent Access classFull price, optimized for real-time alarming vs half price, optimized for less time-sensitive logs, now also supporting full Logs Insights querying

Flashcards — 15 Cards

Click card to flip. Mark right or wrong to track score.

Click to reveal answer
1 / 15
Mark:   Score: 0/0

Practice Quiz — 8 Questions

SCS-C02 scenario style, Easy → Specialty. Select an answer to reveal the explanation.

out of 8 correct