Explain security concepts and controls.
The vocabulary the rest of the exam assumes: defense in depth and the principles that sit under it (CIA, AAA, non-repudiation, Zero Trust, least privilege), then the two ways every security control is classified: by category (how it is implemented) and by type (what it is meant to do).
At 02:10 on a Tuesday, Microsoft Sentinel pages Jordan Avery, security administrator at Ridgeline Health Partners. A night-shift nurse at the Olympia clinic opened an invoice attachment that slipped past the email filter. The macro ran, and Defender for Endpoint saw Word spawn PowerShell, killed the process and isolated the workstation. The nurse’s account had no local administrator rights, the clinic VLAN cannot reach the ChartWell database servers directly, and the one credential the malware did capture was useless without a second factor. By 02:40 Marcus Bell, the SOC lead, closes the incident as contained: one workstation to reimage, no patient data touched.
Nothing about that night was luck. Two controls failed (awareness training and the email filter), and the ones behind them did not. That is the whole of objective 1.1 in one incident: layers of controls, each built on a small set of principles, each classifiable by how it is implemented and what it is supposed to accomplish. The next morning Priya Nair, the CISO, asks Jordan to “map what worked to our control framework,” which is exactly the skill this objective tests.
Defense in depth
Defense in depth is the practice of placing multiple, overlapping controls between an attacker and an asset so that the failure of any single control does not result in compromise. Every control has a failure rate. Filters miss messages, people click, patches arrive late. If the layers fail independently, an attacker has to beat all of them in sequence, and each one that holds also buys time for detection and response.
SY0-701 did not list defense in depth as an objective bullet. SY0-801 makes it the heading that CIA, AAA, non-repudiation, Zero Trust principles and least privilege sit under, and it pulls least privilege forward from the mitigation and access control objectives. Two wording changes: the control type is now written “Deterring” (701 said “Deterrent”), and the categories carry paired names (Technical/logical, Managerial/administrative, Physical/environmental). Zero Trust is cut down to principles here; the control plane and data plane architecture is taught in 3.2.
What makes layering work
Layers only help if they fail independently. Two antivirus products fed by the same signatures, or a firewall and a VPN gateway that both trust the same compromised administrator account, are really one layer. So you build in diversity: mix control categories (a policy, a technology, a locked door, a routine check), mix control types (prevent, detect, correct), and mix vendors where one flaw would otherwise open every door. Layers are people and process as well as technology: awareness training, change control (1.2) and tested backups (3.4) all count. And every layer needs detection, because a layer that cannot report its own failure hides the breach from the ones behind it. This is the “Swiss cheese” model: each control is a slice with holes, and harm occurs only when the holes line up.
Confidentiality, integrity, and availability (CIA)
The CIA triad names the three properties security exists to protect.
| Property | Meaning | Typical mechanisms | Ridgeline failure example |
|---|---|---|---|
| Confidentiality | Information is disclosed only to authorized people, processes and systems. | Encryption at rest and in transit, access control lists, data classification, masking, MFA | A billing export with 9,000 patient records is emailed to the wrong Corvid Billing contact. |
| Integrity | Data and systems are changed only in authorized ways, and unauthorized change is detectable. | Hashing, digital signatures, file integrity monitoring, input validation, version control | An infusion pump drug library is altered so a dose limit is ten times too high. |
| Availability | Authorized users can reach systems and data when they need them. | Redundancy, backups, clustering, load balancing, UPS and generators, DDoS protection | Ransomware encrypts the ChartWell application servers and clinics revert to paper charts. |
A cryptographic hash is a fixed-length fingerprint of a file; change one bit and the fingerprint changes completely (the algorithms are covered in 1.3). Before Elena Ruiz’s team installs a ChartWell update, they compare the hash of the download with the value the vendor published:
A hash on its own proves the file has not changed since the hash was computed. It does not prove who produced it; that gap is closed by a digital signature (see non-repudiation below).
Availability is usually expressed as a percentage of time in service. A year has 8,760 hours, so each extra “nine” cuts allowed downtime by a factor of ten: 99% allows 87.6 hours a year (8,760 × 0.01), 99.9% allows 8.76 hours, 99.99% about 52.6 minutes and 99.999% (“five nines”) about 5.26 minutes.
The three properties pull against each other. Encrypting every disk protects confidentiality, and a lost key then destroys availability. In a hospital, availability of the medication record at the bedside can matter more in the moment than perfect confidentiality, which is why emergency “break the glass” access to patient charts exists and is heavily logged. Security decisions are trade-offs among the three, made according to what the business can least afford to lose.
Expect short situations that ask which property was primarily affected or which a control primarily protects. Work from the verb. Read, copied, leaked, intercepted, exposed: confidentiality. Altered, tampered, injected, corrupted, spoofed record: integrity. Offline, unreachable, encrypted by ransomware, deleted, flooded: availability. Classic traps: hashing is integrity, not confidentiality; encryption is confidentiality first; ransomware is availability first, and becomes a confidentiality incident too only when the stem says data was also stolen.
Authentication, authorization, accounting (AAA)
AAA is the three-step framework behind every controlled access decision. The steps always happen in the same order, and each answers a different question.
- Authentication: who are you? A claimed identity (identification: a username, a certificate subject, a device ID) is proven with one or more factors: something you know (password, PIN), something you have (FIDO2 security key, authenticator app, smart card), something you are (fingerprint, face). Two or more different kinds make multifactor authentication. Systems authenticate too: a laptop proves itself to the network with a device certificate, a server to your browser with a TLS certificate.
- Authorization: what are you allowed to do? Once identity is established, policy decides which resources and actions are permitted. The decision can be based on role, group, attributes such as device health, or explicit rules (the models are taught in 4.5).
- Accounting: what did you do? The system records the activity: sign-in time, source address, resources touched, commands run, data volume, sign-out. Accounting makes investigation and non-repudiation possible; you will also see it called auditing.
The order matters. A request that fails authentication is never evaluated for authorization, and a user can authenticate perfectly and still be denied because policy does not allow the action. Accounting records both outcomes, including the failures.
Two protocols carry AAA for network access and device administration. RADIUS (Remote Authentication Dial-In User Service) and TACACS+ (Terminal Access Controller Access-Control System Plus) both centralize the three functions on a server, so network devices do not each hold their own user list.
| Feature | RADIUS | TACACS+ |
|---|---|---|
| Transport | UDP 1812 (authentication and authorization), UDP 1813 (accounting) | TCP 49 |
| AAA separation | Authentication and authorization are combined in one exchange; accounting is separate | All three are separate exchanges, so each command can be authorized individually |
| Protection | Only the password attribute is obscured; the rest of the packet is readable unless wrapped in TLS (RadSec) or IPsec | The whole packet body is encrypted (strictly, MD5-based obfuscation; newer deployments add TLS 1.3) |
| Typical use | Network access: 802.1X wired and wireless, VPN | Device administration: who may run which commands on routers, switches and firewalls |
At Ridgeline, Cisco ISE answers RADIUS for the 802.1X ports and wireless, and TACACS+ for Tom Okafor’s network team logging in to switches. One session looks like this in the ISE log when Tom changes a trunk port:
Line 1 is authentication (who, from where, how, result). Line 2 is authorization (the command checked against a profile). Line 3 is accounting (what was actually done). In the highlighted line a Northgate Managed IT help desk technician, already authenticated under a named account, asked to wipe the switch configuration and authorization said no. The denial is still recorded.
Non-repudiation
Non-repudiation means a party cannot credibly deny having performed an action, such as sending a message, approving a change or signing an order, because there is evidence that binds the action to that party and that a third party can verify. It goes one step beyond authentication. Authentication convinces the system at sign-in time that you are who you claim. Non-repudiation leaves evidence that convinces someone else, later.
The standard technical mechanism is the digital signature (details in 1.3). The signer hashes the document and runs the signing operation over that hash with a private key that only the signer holds (loosely, “encrypting the hash”). Anyone can check the signature with the matching public key, which a certificate ties to the signer’s identity. A valid signature proves two things at once: the content has not changed since signing (integrity), and it was signed with that specific private key (origin). Because only one party holds the private key, that party cannot plausibly say “it was not me.”
Contrast that with a shared secret. A message authentication code (HMAC) computed with a key that both sender and receiver know proves integrity and that one of the two produced it, and it cannot prove which. Symmetric keys therefore give authentication and integrity, and they do not give non-repudiation. The same logic applies to shared accounts: any of a dozen people could have typed the password, so every log entry is deniable.
Non-repudiation depends on the whole evidence chain:
- Unique identities. One person, one account; one service, one identity. No shared credentials.
- Protected private keys. Keep keys on a smart card, security key or HSM; if a private key is stolen, the owner can credibly say someone else signed.
- Trustworthy time. Systems synchronized to a common time source (NTP) and logs stamped in UTC. Clock drift between systems breaks the timeline.
- Tamper-resistant, retained logs. Logs forwarded off the host to the SIEM, stored write-once (immutable), and kept long enough. Entra ID keeps sign-in logs for 30 days on paid licenses, which is why Ridgeline exports them to Sentinel for longer retention.
When a stem says someone “denies” sending, approving or doing something, or asks for “proof of origin,” the answer is non-repudiation, and the control is a digital signature (asymmetric, private key) backed by logging. Distractors to rule out: encryption alone (hides content, proves nothing about the sender), a hash alone (integrity only, anyone can compute it), HMAC or any shared key (either party could have produced it), and a shared account with good logging (the log cannot say which person). The AAA element that supports it is accounting.
Zero Trust principles
Traditional networks granted trust by location: once a device was inside the firewall, most things would talk to it. Zero Trust removes that assumption. No user, device or network segment is trusted by default; every request to every resource is authenticated, authorized against policy and evaluated in context, and that evaluation is repeated throughout the session. NIST SP 800-207 is the reference description. The objective asks for the principles:
| Principle | What it means | At Ridgeline |
|---|---|---|
| Never trust, always verify (verify explicitly) | Authenticate and authorize every request using all available signals: identity, device health, resource sensitivity, behavior. Network location alone earns no trust. | ChartWell access requires MFA and an Intune-compliant device, on the Tacoma headquarters LAN exactly as on hotel Wi-Fi. |
| Least privilege access | Grant access to one resource for one session, with the minimum rights needed, and no more. | Corvid Billing staff reach the billing interface only, never the wider network. |
| Assume breach | Design as if an attacker is already inside: segment, limit blast radius, monitor everything. | Medical device VLANs are microsegmented so one infected pump cannot scan the rest. |
| Continuous verification | Trust is not permanent. Re-evaluate when risk changes: new location, device falls out of compliance, unusual behavior. | A session is revoked when Defender raises the device risk score to high. |
| Policy-driven, per-request decisions | A central policy decides; enforcement points in front of each resource carry the decision out. | Entra ID Conditional Access decides; the application proxy, ISE and firewalls enforce; Sentinel collects. |
Zero Trust is a strategy, not a product, and it does not mean nobody is trusted: trust is earned per request and never implied. Firewalls and VPNs remain as layers; they stop being the only basis for access. How the policy engine, policy administrator and policy enforcement point fit together is architecture and belongs to 3.2.
A strict access policy switched on at 08:00 Monday locks out everyone whose phone was never enrolled and ends in an emergency rollback. In Entra ID, a Conditional Access policy set to Report-only evaluates every sign-in and logs what it would have done without blocking anyone. Jordan ran Ridgeline’s “ChartWell requires compliant device” policy that way for two weeks, found 37 staff who would have failed, fixed their enrollment, then enforced. It is the impact analysis step of change management (1.2) applied to access policy.
Least privilege
Least privilege means every identity (user, administrator, service account, application, device, API key) is granted only the permissions required for its legitimate function, for only as long as they are required. Its purpose is to limit blast radius: whatever an account can do, an attacker who compromises that account can also do, and so can an honest mistake.
In practice:
- Standard accounts for daily work. Administrators keep a separate privileged account and use it only for administrative tasks.
- Scoped roles instead of broad ones. A help desk role can reset passwords in one organizational unit; it is not a Domain Admin. A cloud function can read one bucket; it does not hold
s3:*. - Just-in-time and time-bound elevation. Privilege is checked out for a task and expires. CyberArk releases the ChartWell SQL administrator credential for two hours against an approved ServiceNow change, then rotates it (see 4.5).
- Services and machines too. The backup agent needs read access to the data it backs up, not domain administrator.
- Default deny and regular access reviews. Start from no access and add. People change jobs and keep old rights (privilege creep); periodic recertification removes them.
In the cloud the same idea is written as policy. Sam Whitaker’s team found that the MyRidgeline document-upload function had been deployed with a wildcard. Ninety days of AWS CloudTrail activity showed it used two actions on one bucket, so the Terraform now grants exactly those:
Two relatives appear next to it in questions. Need to know applies the same idea to information: clearance alone is not enough, you must need that specific data for your task. Separation of duties splits a sensitive process so that no one person can complete it alone. Least privilege limits what one identity can do; separation of duties limits what one person can finish.
If the identity provider or MFA service fails, administrators may be unable to sign in to fix it, so organizations keep a few emergency access (“break-glass”) accounts with standing high privilege. Microsoft’s guidance for Entra ID is at least two cloud-only accounts, excluded from the Conditional Access policies that could lock them out, protected with phishing-resistant authentication such as FIDO2 security keys stored securely, with an alert on every sign-in and a scheduled test. A documented, monitored exception is consistent with least privilege; an undocumented standing admin account violates it.
Control categories
A security control is any safeguard or countermeasure that reduces risk to the confidentiality, integrity or availability of an asset. CompTIA classifies controls on two independent axes. The category answers “how is this control implemented, and by what?” The type answers “what is it meant to accomplish?” Every control has one category that best describes it and one or more types.
| Category | Implemented by | Examples | Ridgeline |
|---|---|---|---|
| Technical/logical | Hardware, software or firmware. The system enforces it without a person acting each time. | Firewall rules, encryption, access control lists, MFA, EDR, IDS/IPS, group policy, DLP | Palo Alto rule denying clinic VLANs to SQL port 1433; BitLocker on laptops |
| Managerial/administrative | Management decisions and documents that direct and oversee the security program. | Security policies, risk assessments, security planning, vendor risk reviews, control selection | Priya’s annual risk assessment; the acceptable use policy |
| Physical/environmental | Tangible barriers and facility systems that protect buildings, rooms, equipment and people. | Locks, fences, bollards, badge readers, access control vestibules, guards, cameras, lighting; fire suppression, HVAC, temperature and humidity sensors, water detection, UPS | Victor Hale’s badge-controlled data hall in Tacoma with clean-agent fire suppression and leak sensors |
| Operational | People carrying out day-to-day procedures. | Awareness training, access reviews, log review, backup and restore runs, patch routines, change control execution, incident response, logged patrol rounds | Marcus’s SOC triage shift; quarterly ChartWell access recertification |
Telling the categories apart
“Logical” is the older word for technical: controls that live in code and configuration. The test is whether technology enforces the outcome. A password policy document is managerial; the domain setting that rejects a 6-character password is technical. Managerial controls set direction (see 5.1 and 5.2) and work only through the controls they mandate. For physical/environmental: if you can touch it, or it protects something you can touch, it is physical, and the “environmental” half is a reminder that heat, water, smoke and power loss take systems down as surely as an intruder, so cooling, fire suppression and power conditioning are security controls protecting availability. Operational controls are executed by people following procedures.
One activity shows the split: management writes the backup policy (managerial), the backup software runs the encrypted job (technical), the tapes sit in a locked, climate-controlled vault (physical/environmental), and an operator performs the monthly restore test (operational).
Other frameworks slice differently (the HIPAA Security Rule uses administrative, physical and technical safeguards), and some controls sit on a border: a security guard is a person performing a duty (operational) whose purpose is protecting a facility (physical), and references disagree. This guide treats a posted guard as physical (deterring and preventive) and the patrol procedure as operational. When a question forces a choice, read the stem’s emphasis: a patrol schedule and logged rounds describe a procedure; fencing, lighting and guards describe physical security.
The most common error on this objective is answering a category question with a type, or the reverse. If the options are technical, managerial, physical, operational, the question is about implementation. If they are preventive, detective, corrective and so on, it is about purpose. When a stem asks for both, settle each axis separately. Favorite pairs to keep straight: a policy is managerial, the training that teaches the policy is operational, the system setting that enforces it is technical.
Control types
The type describes the control’s function relative to an unwanted event: does it act before, during or after, and what does it do? A control can serve more than one function, so choose the type that matches what the stem says the control is for.
Preventive
A preventive control stops the unwanted event from occurring. It acts before the fact and removes the opportunity: a firewall rule drops the packet, MFA rejects the stolen password, file permissions refuse the read, a locked door stays shut, application allow-listing refuses to run the unknown binary. Hardening, patching and encryption are preventive.
Deterring
A deterring control discourages someone from attempting the action by raising the perceived chance of being caught or punished. It works on the attacker’s decision, and it does not physically or technically block anything. Examples: warning signs, visible cameras, lighting, a uniformed guard at the entrance, a login banner stating that activity is monitored and unauthorized use will be prosecuted, a published sanctions policy. The test: if the attacker ignores it, does anything stop them? If nothing does, it is deterring and not preventive.
Corrective
A corrective control acts after an event to limit the damage and return the system to a known good state. Restoring from backup, reimaging an infected workstation, applying the patch that closes the exploited hole, quarantining a malicious file and failing over to the Spokane site are all corrective.
Detective
A detective control identifies and records an event while it is happening or after it has happened. It does not stop anything by itself; its value is turning an unknown compromise into a known one quickly. Examples: SIEM correlation rules, intrusion detection systems, log review, file integrity monitoring, audits and access reviews that find excess rights, motion sensors, vulnerability scans. Detective controls are paired with corrective ones, because detection without response changes nothing.
Compensating
A compensating control is an alternative safeguard used when the normally required (primary) control cannot be implemented for a legitimate technical or business reason. It must address the same risk to a comparable degree; it is a substitute, not an excuse to skip the requirement.
Ridgeline’s clearest example is its medical devices. Policy requires supported, patched operating systems. Many of its roughly 1,100 networked medical devices run vendor-locked OS versions that Ridgeline cannot patch itself: the manufacturer validates the software on a regulated device, and for these models it has released no supported update. The compensating set: each device class is isolated in its own VLAN with firewall rules permitting only the specific servers and ports it needs, Cisco ISE profiles the devices and blocks anything else on those ports, and traffic is monitored for anomalies. PCI DSS formalizes the same idea: an organization that cannot meet a stated requirement may document a compensating control that meets the intent and rigor of the original requirement.
Directive
A directive control tells people what they must or must not do. It guides behavior toward the desired outcome through instruction: policies, standards and procedures, an acceptable use policy that staff sign, mandatory training assignments, and signs such as “Authorized personnel only” or “Badge must be visible at all times.” It depends on people complying. Compare the sign pair: “All visitors must sign in at reception” directs; “Trespassers will be prosecuted” deters.
One control, several types. A visible camera deters; its recording detects. A guard deters by being seen and prevents by refusing entry. An intrusion detection system is detective; put the same engine inline as an intrusion prevention system and it is preventive. Endpoint protection that spots malware and quarantines it is detective and corrective in one product. The stem decides.
Putting both axes together
Because category and type are independent, any control can be placed in a grid. An audit finding such as “no detective control over privileged access” describes an empty cell.
| Type | Technical/logical | Managerial/administrative | Physical/environmental | Operational |
|---|---|---|---|---|
| Preventive | Firewall deny rule, MFA, disk encryption | Pre-employment background check requirement; vendor onboarding standard | Door lock, access control vestibule, bollards | Change approval performed before deployment; visitor escort |
| Deterring | Login warning banner | Published sanctions for policy violations | Warning signs, visible cameras, lighting | Publicized random bag checks and patrol rounds |
| Detective | Sentinel alert rules, IDS, file integrity monitoring | Internal audit program, risk assessment findings | Motion sensors, door-held-open alarms, smoke and water detectors | Daily log review, quarterly access recertification |
| Corrective | EDR quarantine and host isolation, automated rollback | Corrective action plan after an audit | Fire suppression discharge, generator start on power loss | Restoring from backup, reimaging, incident response |
| Compensating | VLAN isolation of unpatchable pumps | Exception standard that mandates extra management review while a requirement is unmet | Locked cabinet where a dedicated room is impossible | Manual dual sign-off where the system cannot enforce separation of duties |
| Directive | Pop-up reminder to classify a document | Acceptable use policy, security standards | “Authorized personnel only” signs | Standard operating procedures, assigned training |
Go back to the 02:10 incident with this vocabulary. Awareness training: operational, preventive (failed). Email filter: technical, preventive (failed). Defender behavioral rule: technical, detective; its automatic isolation: technical, corrective. No local admin rights and MFA: technical, preventive, both expressions of least privilege and Zero Trust. VLAN segmentation: technical, preventive. Reimaging the workstation: operational, corrective. The acceptable use policy the nurse signed: managerial, directive. That one-paragraph mapping is what Priya asked Jordan for.
Scenario workshop
Situation. A Tenable scan flags 140 infusion pumps running an operating system that no longer receives security updates. The manufacturer confirms that no supported upgrade exists for this model and that it will not support a pump whose validated software the hospital has modified. Ridgeline policy requires all networked systems to run supported, patched software. Dr. Alan Reyes will not accept removing the pumps from the network because charting depends on it.
Question. What kind of control does Jordan need, and what should it look like?
Answer and reasoning. A compensating control: the required control (patching) cannot be applied for a legitimate reason, so an alternative must cover the same risk (exploitation of known vulnerabilities over the network). Place the pumps in a dedicated VLAN, allow only the pump server’s address and required ports through the firewall, have ISE refuse any other device type on those switch ports, and monitor the segment in Sentinel. Document the exception, its owner and an expiry tied to device replacement. Calling this corrective is tempting and wrong: nothing has gone wrong yet that is being repaired. A written policy exemption alone is managerial paperwork that leaves the risk untouched. Disconnecting the pumps would prevent the attack and sacrifice availability and patient safety, which the business has ruled out.
Situation. A firewall rule opening the ChartWell database subnet to a Northgate Managed IT address range was added at 23:40 on a Saturday. The Palo Alto audit log shows the change was made by the account fw-admin, a local administrator account whose password is stored in a team spreadsheet used by four Ridgeline engineers and two Northgate staff. Everyone asked denies making the change, and the firewall’s clock is nine minutes ahead of the VPN concentrator’s.
Question. Which security concept has failed, and what fixes it?
Answer and reasoning. Non-repudiation has failed. The log faithfully recorded an account, and six people share that account, so each can credibly deny the action, and the clock drift makes it hard to match the change to any one VPN session. The fix is individual named accounts for firewall administration authenticated centrally through TACACS+ with MFA, the shared local account vaulted in CyberArk for emergencies with every checkout recorded, all devices synchronized to the same NTP source, and logs forwarded to Sentinel. Confidentiality is tempting because a database was exposed, and that is the consequence, not the concept that broke. Authentication did not fail either: the password was correct. More logging changes nothing while the identity in the log is shared.
Knowledge check20 questions
Tap an answer to check it instantly and read the explanation.
1. During a firmware audit, a biomedical engineer finds that the drug library on several infusion pumps was modified so that the maximum dose limit for one medication is ten times higher than the approved value. No data left the network and the pumps stayed online. Which security property was PRIMARILY violated?
2. Before installing a vendor update, an administrator runs Get-FileHash on the downloaded installer and confirms the SHA-256 value matches the one printed in the vendor release notes. What does this check establish on its own?
3. A help desk technician signs in to a core switch with a valid username, password and MFA prompt. When the technician enters a command to erase the configuration, the switch rejects it, and the attempt appears in the central log. Which AAA function rejected the command?
4. A network team wants central control over which engineers can run which commands on routers and firewalls. They require each command to be authorized individually and the entire protocol payload to be encrypted in transit. Which choice BEST meets these requirements?
5. A purchasing manager claims they never approved a large order, although the approval record exists. The company wants future approvals to provide evidence that a third party could verify and that the approver could not credibly deny. Which mechanism BEST provides this?
6. Which statement BEST describes a Zero Trust principle?
7. An audit finds that the service account used by a backup agent is a member of Domain Admins because that made the original installation easier. The agent only needs to read the file shares it backs up. Which principle is being violated?
8. A CISO completes the annual enterprise risk assessment and uses the results to decide which security controls the organization will fund next year. Which control category does the risk assessment belong to?
9. Before the password prompt, a jump host displays: 'Access is restricted to authorized users. All activity is logged and monitored. Unauthorized use may result in prosecution.' The banner itself does not block any connection. Which control type BEST describes the banner's warning of monitoring and prosecution?
10. New staff must read and sign the acceptable use policy, and a sign at the data hall entrance reads 'Badge must be visible at all times.' Neither measure blocks anything or threatens a penalty. Which control type do both represent?
11. Which design BEST demonstrates defense in depth for a database that holds patient records?
12. A Tenable scan at Ridgeline flags 140 infusion pumps running an operating system that no longer receives security updates. The manufacturer confirms no supported upgrade exists and that it will not support pumps whose validated software has been modified. Ridgeline policy requires supported, patched software on all networked systems, and the chief medical information officer states the pumps must stay networked for charting. Replacement is two budget cycles away. Jordan proposes isolating the pumps in a dedicated VLAN, permitting traffic only to the pump server on the required ports, and monitoring the segment in the SIEM. Which control type is Jordan proposing?
13. Ridgeline plans to require an Intune-compliant device for all remote access to ChartWell, and Priya Nair wants no surprise lockouts on go-live day. Jordan creates a Conditional Access policy named 'ChartWell requires compliant device' and sets it to report-only. The next day Jordan reviews this sign-in record: User: d.cho | Application: ChartWell Remote | Authentication details: Password succeeded, Authenticator push succeeded | Device: not compliant | Conditional Access (report-only): Failure | Sign-in status: Success. What does the record tell Jordan?
14. A firewall rule exposing the ChartWell database subnet to an external address range was added at 23:40 on a Saturday. The audit log entry reads: admin=fw-admin | action=add rule 'tmp-db-access' | source=VPN pool. The fw-admin account is a local account whose password is kept in a spreadsheet used by four Ridgeline engineers and two Northgate Managed IT staff. All six deny making the change. The firewall clock is nine minutes ahead of the VPN concentrator. Which change would BEST address the underlying weakness?
15. Ridgeline's contract with a hosting provider for its clinic scheduling system specifies 99.9% availability measured over a calendar year (8,760 hours). During the year the system had three unplanned outages lasting 3 hours, 4 hours and 2.5 hours. No data was lost or disclosed in any outage. The provider's annual report describes the year as 'high availability with only three brief interruptions', and Grace Lindqvist asks Jordan whether the contractual availability commitment was actually met for the year. Which statement is correct?
16. Corvid Billing staff reach Ridgeline through a site-to-site VPN that terminates in a subnet labeled 'trusted partners'. The firewall rule for that subnet is: source=partner subnet | destination=all server VLANs | service=any | action=allow. Corvid reports that an analyst's laptop was under an attacker's control for six hours after a phishing email. Corvid staff only need the billing interface on one server. Priya Nair asks for the design change that BEST applies Zero Trust and least privilege. What should Jordan recommend?
17. Every quarter, each Ridgeline department manager receives a list of staff with access to ChartWell and must confirm or revoke each entry by following a documented procedure. In the last cycle the review found 46 accounts that still had access from previous job roles; those were removed afterwards under a separate ticket. An auditor building Ridgeline's control matrix asks Jordan how this quarterly review should be recorded. Considering only the review activity itself, which category and type BEST describe the control?
18. An assessor reviewing a regional credit union lists the controls that protect its online banking database: a perimeter firewall with a default deny rule, a web application firewall, TLS for all connections, MFA for customers, and full disk encryption on the database servers. No alerts are generated and no logs are reviewed. Last year an attacker used a stolen administrator session to copy records for three weeks before a customer complaint revealed the breach. Which addition would BEST strengthen the layered defense?
19. Sam Whitaker reviews the IAM policy attached to the MyRidgeline document-upload function: Effect: Allow | Action: s3:* | Resource: *. CloudTrail logs for the past 90 days show the function has only ever called s3:PutObject and s3:GetObject on the bucket myridgeline-uploads-prod. The function stores insurance card images that patients upload through the portal, and a recent penetration test noted that a flaw in the function's code would let an attacker act with whatever permissions the function holds. Which change BEST applies least privilege without breaking the function?
20. Marcus Bell summarizes an incident at the Olympia clinic for the post-incident review: (1) A login banner warned that the workstation was monitored. (2) Defender for Endpoint raised an alert when Word started PowerShell. (3) Defender automatically isolated the workstation from the network. (4) The email filter failed to block the attachment. (5) The help desk reimaged the workstation from the standard build. The nurse's account had no local administrator rights and no patient data was touched. Which two entries describe corrective controls in action?