CompTIA CySA+ Cheat Sheet: Key Facts
Every key takeaway and common trap from the CompTIA CySA+ study guide on one page, grouped by chapter. Tap a lesson link to read the full explanation.
70 facts · 11 chaptersFacts checked 2026-11-15
Chapter 1
Architecture & Log Sources
Key takeaways
- Flow records, firewall logs and DNS logs describe connections and names, so naming the user or the process requires endpoint or identity telemetry. → What each log source can and cannot prove
- Correlation across sources is only valid after every timestamp is placed on one synchronized reference clock, normally UTC. → Time synchronization and log normalization
- Silence in a log means nothing until you confirm the source was placed, configured and retained to see that activity. → Sensor placement, logging levels and retention
- Containers and serverless functions leave no host to image, so their evidence exists only if it was shipped off while they ran. → Cloud, containers, serverless and virtual machines
- With single sign-on, the identity provider holds the authentication record and the application holds the record of what the session did. → Identity, directory and federation logs
Common trap
- Watch for answers that turn an empty result into a clean bill of health. No entries from an edge sensor, a deny-only firewall, an expired retention window or the wrong identity system means the evidence was never collected there, not that the activity did not occur. → Sensor placement, logging levels and retention
Quick reference
| Source | What it records | What it cannot tell you |
|---|---|---|
| Flow records | Hosts, ports, protocol, bytes, duration | User, process or payload |
| Firewall without a proxy | Allowed or denied connections | Full URLs requested |
| DNS resolver log | Names queried and answers | The page path or the requesting program |
| Endpoint agent | Process, parent, command line, logon session, sockets | Hosts where it is not installed |
Chapter 2
Indicators of Malicious Activity
Key takeaways
- Long random-looking subdomains under one domain point to DNS tunneling, while a burst of failed lookups ending in one success points to a domain generation algorithm. → Network indicators: DNS abuse, odd protocols and unusual volume
- A process is authenticated by its image path and its parent, never by its name alone. → Host indicators: process lineage, masquerading and built-in tools
- Order matters: disabling auditing, then clearing logs, then stopping logging is a deliberate attempt to destroy evidence. → Persistence and anti-forensics on hosts
- Many failures on one account is brute-force guessing, while one password tried across many accounts is password spraying. → Identity and application indicators
- Scanning from an inventoried assessment host inside a published window, or impossible travel through the company VPN, is benign once the source is verified. → Network indicators: DNS abuse, odd protocols and unusual volume
Common trap
- Plausible names are the bait. A service called something like a disk or update helper, or a process named svchost.exe, sounds routine on purpose. Choose the answer that checks the path, the parent, the account and the change record, not the one that trusts the name. → Persistence and anti-forensics on hosts
- Impossible travel is an inference from addresses. If the far-away address is the company's own VPN egress range, the device is the same and the pattern matches the user's baseline, it is a false positive to verify and close, not an account takeover. → Identity and application indicators
Quick reference
| Pattern | Most likely meaning |
|---|---|
| Thousands of long, high-entropy subdomains of one domain, similar-size TXT answers, no connection to the resolved address | DNS tunneling for command and control or data |
| Hundreds of never-seen names, nearly all NXDOMAIN, then one resolves and a session opens to it | Domain generation algorithm malware |
| Port 443 sessions whose first bytes are an SSH banner and no TLS ClientHello | Protocol tunneled over an allowed port |
| Tens of gigabytes leaving a host whose role never sends data out | Data exfiltration |
Chapter 3
Investigation Tools & Threat Hunting
Key takeaways
- Full packet capture holds payload but is kept briefly, while flow records keep months of metadata and never the content. → Packet capture, flow data and network monitoring sensors
- A not-found reputation result only means nobody submitted that exact file before, so local analysis continues. → Endpoint telemetry, file reputation and sandboxes
- In an email, only the Received lines your own servers wrote, from the top down to your inbound gateway's hop, can be trusted. → SIEM correlation rules and email header analysis
- Behaviors and techniques stay useful long after addresses, domains and hashes have rotated. → Judging and sharing threat intelligence
- A hunt needs a testable hypothesis, a negative result is bounded by coverage, and success ends in a tuned automated detection. → Running a threat hunt from hypothesis to detection
Common trap
- Bulk-blocking every address in a report feels decisive, which is why it is a favorite distractor. Old or low-confidence indicators on shared infrastructure break legitimate services and expire quickly; the behavior described in the report is what still detects the adversary next month. → Judging and sharing threat intelligence
Quick reference
| Tool or data | What it keeps | Typical use |
|---|---|---|
| Full packet capture | Every byte, including payload | Reassemble a stream in a protocol analyzer to read exactly what was sent |
| Flow records | Addresses, ports, times, duration, direction, byte counts | Prove a conversation happened months later |
| Network security monitoring sensor | One metadata log per connection, with fields such as the TLS server name | Searchable history of protocol details without storing payload |
| Signature-based IDS | Only traffic that matched a rule | Alert on known patterns |
Chapter 4
Process Improvement, Automation & AI
Key takeaways
- Enrichment is the best first automation because it gathers facts and leaves the decision to the analyst. → AI governance: accountability, audit trails and validation
- A playbook holds decisions, roles and escalation, while a runbook holds the exact technical steps and commands. → Standard processes: playbooks, runbooks and the case record
- Tune a noisy detection by narrowing it instead of disabling it, because disabling removes coverage for every other account. → Integrations, alert tuning and measuring improvement
- AI output is an input to an analyst, so every claim must be found in the real evidence before anyone acts on it. → AI in security operations: good uses and main risks
- Governance means a named person owns each disposition and the AI input, output and verified evidence are recorded. → AI governance: accountability, audit trails and validation
Common trap
- Asking the same model again, or having a second model agree, is not verification. A model can be confidently wrong twice in the same way. Verification means finding the cited events in the real host and network telemetry. → AI in security operations: good uses and main risks
Quick reference
| Document | What belongs in it | Example item |
|---|---|---|
| Playbook | Decision points, who holds authority, escalation and notification order | Who may approve isolating a production server |
| Runbook | Exact commands, queries and system steps, with expected output | How to capture a memory image on an affected host |
| Risk | What it looks like | Correct response |
|---|---|---|
| Hallucination | Fluent output naming a technique, key or address that is not in the evidence | Keep it out of the case and check every claim in the raw data |
| Prompt injection | Hidden text in a reported email tells the model to call it safe | Treat the output as untrusted and judge headers, links and attachments |
| Data exposure | Case notes with customer names sent to an outside service | First confirm retention, reuse for training and who can read it |
| Poisoning | Mass-closed true alerts become training labels and the model learns to ignore them | Review and sample dispositions before they are used to retrain |
Chapter 5
Vulnerability Scanning Methods
Key takeaways
- A scan only speaks about the assets in its scope, so inventory and discovery come before any remediation percentage means anything. → Asset inventory, discovery scans and scan scope
- Credentialed scans and agents read installed software and settings from inside the host, which removes banner guesses and reveals client-side software. → Credentialed, non-credentialed and agent-based scanning
- Silence from a host that is off, filtered or blocked is missing data, not a clean result. → Scanner placement: internal, external and what sits in between
- Fragile operational technology and medical devices are assessed passively and through vendor information, with segmentation covering the residual risk. → Passive scanning, fragile devices and scan scheduling
- Web applications, container images, infrastructure-as-code and hardening baselines each need their own kind of scan in the right place. → Web applications, containers, infrastructure-as-code and baselines
Common trap
- Wrong answers often claim that a faster scan, a longer timeout or more retries will turn potential findings into confirmed ones or reach a laptop that is not on the network. Speed and timeouts change neither the evidence nor the reachability; only authenticated or agent-based collection does. → Credentialed, non-credentialed and agent-based scanning
Quick reference
| Target | What the scan needs | What it misses otherwise |
|---|---|---|
| Web application | A maintained session: credentials, a recorded login sequence and the logout link excluded | Everything behind the login, such as accounts, payments and administration |
| Container | Image scanning in the build pipeline and the registry | Libraries inside image layers, which a host scan of the node never reads |
| Infrastructure-as-code | Template scanning before apply plus assessment of the deployed cloud state | Drift, such as a setting changed by hand in the provider console |
| Hardening baseline | A credentialed configuration audit policy run against the adopted benchmark | Per-setting compliance evidence that a vulnerability policy never produces |
Chapter 6
Analyzing Scan Output
Key takeaways
- A finding inferred from a banner or response header is a hypothesis until the installed version is confirmed on the host. → Validating findings and spotting false positives
- False negatives are worse than false positives because a missed exposure never enters the queue; failed credentials and stale plugin feeds are common causes. → False negatives, failed credentials and stale plugin feeds
- Present, reachable and exploited are three separate judgments, and a compensating control never turns a real finding into a false positive. → Present, reachable, mitigated or already compromised
- Missing patches, misconfigurations, end-of-life platforms and shared templates or images each point to a different remediation. → Reading the kind of finding: patch, configuration, end of life or shared source
- Consolidate detections to unique vulnerability-per-asset pairs, but keep separate code instances of the same weakness class as separate findings. → Consolidating and deduplicating findings
Common trap
- Some output describes an event rather than a weakness. An unlinked file in an upload directory that runs supplied commands and was modified weeks ago is working attacker tooling. Deleting it and closing a ticket destroys evidence; the right path is to preserve the file and host evidence and invoke incident response. → Present, reachable, mitigated or already compromised
Quick reference
| Kind of finding | Clue in the output | Where remediation belongs |
|---|---|---|
| Missing patch | Installed version is older than a fixed release | Apply the update, then rescan |
| Misconfiguration | Software is current but offers deprecated TLS versions or weak ciphers | Hardening change in the server's own configuration, then a verification rescan |
| End of life | Vendor support ended and no patch is listed | Upgrade or replace the platform, isolating it in the meantime |
| Shared source | Host built from a template, a container image or an infrastructure-as-code definition | Fix the source, rebuild, then scan everything built from it |
Chapter 7
Prioritization, Mitigation & Controls
Key takeaways
- A severity score measures how bad a flaw would be if exploited, not the risk it poses in your environment. → Severity scores versus risk, and how CVSS v4.0 is organized
- Confirmed exploitation, internet reachability and asset criticality move a finding up the queue, and a change in exposure is a reason to re-rank it. → Severity scores versus risk, and how CVSS v4.0 is organized
- A finding closes on verification by rescan, not on a deployment tool's success report. → Remediation paths, change control and verifying the fix
- Compensating controls, risk transfer and segmentation reduce or shift exposure but leave the defect open and tracked. → Compensating controls, exceptions, acceptance and transfer
- Classify a control's type by what implements it and its function by what it does: prevent, detect or correct. → Classifying controls by type and by function
Common trap
- The most common wrong answer picks the finding with the highest label: a critical flaw on an air-gapped test bench or a powered-off spare server. A medium-rated authentication bypass on an internet-facing customer portal with exploitation in the wild outranks both. → Severity scores versus risk, and how CVSS v4.0 is organized
- When a register lists a confirmed patch, a removed service, a filtering rule waiting on the patch and a documented decision to do nothing, only the filtering rule is the compensating control. Patching and removal are remediation because the vulnerable code is fixed or gone, and a decision to take no action is risk acceptance, not a control at all. → Classifying controls by type and by function
Quick reference
| Classification | Category | Example |
|---|---|---|
| Type | Administrative, also called managerial | A written standard requiring quarterly review of server images |
| Type | Technical | A configuration baseline enforced by an agent |
| Type | Physical | A locked rack cabinet with a camera over it |
| Function | Preventive | A firewall rule blocking the vulnerable port |
| Function | Detective | An alert that fires when that port is probed |
| Function | Corrective or responsive | A documented rebuild procedure if the host is compromised |
Chapter 8
Attack Frameworks & the Response Process
Key takeaways
- Use MITRE ATT&CK for behaviors and detection coverage, the Cyber Kill Chain for where a control breaks a campaign earliest, and the Diamond Model for pivoting and linking intrusions. → Three attack frameworks, three different questions
- In ATT&CK the tactic is the goal, the technique is the method, and the procedure is the specific way one actor carried it out. → Using MITRE ATT&CK: tactics, techniques and procedures
- Detection surfaces a possible incident, while analysis confirms it, scopes it, sets severity and declares it. → The incident response lifecycle, old and new framing
- Scope before containment, root cause before eradication is called complete, and eradication before recovery. → Containment, eradication, recovery and post-incident activity
- NIST SP 800-61 Rev. 3 maps incident response to the CSF 2.0 functions, with lessons learned feeding improvement across all of them. → The incident response lifecycle, old and new framing
Common trap
- Wrong options often pair the right model with the wrong job, such as using the kill chain to measure detection coverage or ATT&CK to attribute a group. Ask what the model is built from: stages, a behavior catalog, or relationships. → Three attack frameworks, three different questions
- A distractor often says malware was removed and service restored, so the case is done. If nobody found how the intruder got in, eradication removed only the symptom, and reinfection through the same path is the expected result. → Containment, eradication, recovery and post-incident activity
Quick reference
| Model | Built around | Best question it answers |
|---|---|---|
| MITRE ATT&CK | A catalog of tactics and techniques | Which behaviors do our detections cover or miss? |
| Cyber Kill Chain | Seven ordered stages | Where would one control have stopped this campaign earliest? |
| Diamond Model | Four vertices: adversary, capability, infrastructure, victim | Do two intrusions share an operator, and what else links to this clue? |
Chapter 9
Incident Response Techniques
Key takeaways
- Triage ranks alerts by how likely they are to be real and how much damage they would do, and scope follows the indicator or credential, not the first host. → Triage and scoping an incident
- Collect the most volatile evidence first, because memory, processes and connections vanish at a reboot. → Collecting evidence: order of volatility and safe access
- Hash an image at acquisition, verify the hash after every copy, and log every handoff and responder action as it happens. → Integrity, chain of custody and legal hold
- Isolate a compromised host from the network rather than powering it off, and revoke sessions and tokens when an identity is compromised. → Containment techniques that keep the evidence
- Root cause is the condition that made the compromise possible, and recovery starts from a backup taken before the first attacker activity. → Eradication, recovery and root cause
Common trap
- Root cause questions tempt you to pick the first event on the timeline, such as the employee who opened a booby-trapped invoice. Root cause is the condition that made the compromise possible, such as workstations allowed to run macros from internet mail, because fixing that condition is what stops a repeat. → Eradication, recovery and root cause
Chapter 10
Vulnerability Reporting & Communication
Key takeaways
- A report validates, deduplicates and assigns findings with an action and priority; a raw scanner export is only evidence. → From raw scan export to an actionable finding
- Executives get impact, options, cost and residual risk; owners get hosts, versions, windows and verification; auditors get compliance evidence. → Matching the report to the audience
- Trends, aging by severity, recurrence and scan coverage say more than any single end-of-month count. → Metrics that show direction: trends, aging, recurrence and coverage
- Risk acceptances and exceptions need a named owner with authority, compensating measures, an expiry or review date and the residual risk. → Risk acceptance, exceptions, action plans and escalation
- Credibility depends on honest severity, evidence-based disputes, approved disclosure channels and human verification of AI drafts. → Credibility: honest severity, controlled disclosure and AI-drafted reports
Common trap
- Wrong answers hand a reader the right artifact for someone else: CVE numbers and host lists to the risk committee, financial framing and risk appetite to the engineers, a raw scan export to the auditor. Ask what this reader must decide or do, then pick the content that serves that one action. → Matching the report to the audience
- An AI assistant can draft a monthly narrative that claims hosts were remediated or cites a CVE found in no scan result. Adding a machine-generated label or asking a public model for a second opinion is not the fix. The analyst checks every claim against the scan and ticket data of record and stays accountable for what is issued; pasting vulnerability details into an external model also exposes sensitive data. → Credibility: honest severity, controlled disclosure and AI-drafted reports
Quick reference
| Reader | Artifact | What it contains |
|---|---|---|
| Risk committee or executives | Executive summary | Business exposure, the decision needed about the settlement pause, cost and the risk of deferring |
| Owning platform team | Action plan or remediation ticket | Exact hosts, version to install, agreed window, accountable owner and how the fix is verified |
| Division director | Risk scorecard | Posture over several quarters for that business unit |
| External auditor or assessor | Compliance report | Control status and retained evidence that the required assessment was performed |
Chapter 11
Incident Reporting & Metrics
Key takeaways
- Keep what was observed separate from what was concluded, and give every conclusion a confidence level and the evidence still missing. → Writing a defensible incident record
- Record each response action with who did it, when it took effect and which system it touched. → Writing a defensible incident record
- During an incident, one coordinator sends scheduled updates on a channel the attacker cannot reach. → Communicating while the incident is live
- A post-incident review examines the process, not individuals, and every lesson leaves with an owner, a deliverable and a date. → After-action reports, root cause and lessons learned
- Mean time to detect, respond, remediate and close each measure a different interval, and a timing metric is always read next to a quality measure. → Response metrics and how they mislead
Common trap
- Any timing metric improves if cases are closed before they are properly validated. When response times keep falling while real intrusions turn up closed as false positives, the fix is to read the metric beside the true-positive rate and a sample review of closures. The same applies to the share of alerts closed by automation: it shows workload absorbed, not whether those closures were right. → Response metrics and how they mislead
Quick reference
| What the review found | Kind of failure | Corrective action |
|---|---|---|
| No control was ever deployed | Missing control | Implement it |
| The control ran and missed the activity | Detection miss | Tune it or change the method |
| A correct alert sat unread in the queue | Output not acted on | Fix triage capacity and alert quality |
| Metric | Clock starts | Clock stops |
|---|---|---|
| Mean time to detect | The malicious activity begins | The team knows about it |
| Mean time to respond | The team knows | Effective action against the threat |
| Mean time to remediate | The issue is known | The underlying cause is eliminated |
| Mean time to close | The case opens | Formal closure with documentation |