Free GDSA Practice Test 2026 — GIAC Defensible Security Architect Exam Questions

Master the GIAC Defensible Security Architect (GDSA) exam with free practice questions covering all 8 official GDSA domains. Each question includes a detailed explanation aligned with SANS SEC530 — no signup required.

GDSA Exam Overview

Practice by GDSA Domain

Zero Trust Architecture and Networking (16%)

Free GDSA practice questions on Zero Trust pillars, microsegmentation, software-defined perimeters, identity-aware proxies, and policy enforcement points. Practice this domain →

Layer 1, Layer 2, and Layer 3 Defense (16%)

Free GDSA practice questions on switch security, VLAN hardening, 802.1X/NAC, routing security, ARP/MAC protections, and physical-layer controls. Practice this domain →

Network Defenses, Proxies, Firewalls, and Remote Access (16%)

Free GDSA practice questions on next-gen firewalls, forward and reverse proxies, VPN and remote access, IDS/IPS, and TLS inspection. Practice this domain →

Data Discovery, Governance, Mobility, and Data-Centric Security (14%)

Free GDSA practice questions on data classification, DLP, data governance, encryption at rest and in transit, rights management, and mobile data security. Practice this domain →

Cloud-Based and Hybrid Security Architecture (12%)

Free GDSA practice questions on cloud security controls, hybrid architecture, CASB, cloud IAM, shared responsibility, and cloud network security. Practice this domain →

Fundamental Security Architecture Concepts (10%)

Free GDSA practice questions on defense in depth, threat modeling, security frameworks, risk management, and security design principles. Practice this domain →

Zero Trust Endpoints and Host Hardening (10%)

Free GDSA practice questions on endpoint hardening, EDR, application whitelisting, host firewalls, patch management, and privileged access. Practice this domain →

IPv6 and Modern Network Risks (6%)

Free GDSA practice questions on IPv6 addressing, IPv6 security risks, dual-stack risks, SLAAC and RA Guard, and modern network threats. Practice this domain →

8 Free GDSA Sample Questions with Answers

Each question below includes 4 answer options, the correct answer, and a detailed explanation. These are real questions from the FlashGenius GDSA question bank.

Sample Question 1 — Cloud-Based and Hybrid Security Architecture

A platform team must ensure a container cannot exhaust the host's memory. Which docker run option enforces this restriction using Linux control groups?

  1. A. --cap-drop=ALL
  2. B. --memory=512m (Correct answer)
  3. C. --security-opt no-new-privileges
  4. D. --read-only

Correct answer: B

Explanation: B is correct because the memory limit is applied through the cgroup memory controller, which caps the container's memory usage. A is incorrect because this removes Linux capabilities from the container's processes, which restricts privileges, not resource usage. C is incorrect because this prevents processes from gaining privileges through setuid binaries and does not limit memory. D is incorrect because this mounts the container's root filesystem read-only and does not limit memory.

Sample Question 2 — Data Discovery, Governance, Mobility, and Data-Centric Security

A three-tier HR application connects to its database with a single service account, svc_hrweb. During an investigation, database activity monitoring shows a bulk export of salary data, but every query in the log is attributed to svc_hrweb. Which design change gives the database per-user attribution for future queries?

  1. A. Enable transparent data encryption on the tables that hold salary data
  2. B. Place an inline database firewall between the application tier and the database
  3. C. Rotate the svc_hrweb password regularly and store it in a secrets vault
  4. D. Pass each end user's identity to the database through Kerberos delegation (Correct answer)

Correct answer: D

Explanation: D is correct because when the database authenticates the actual user, activity monitoring and native auditing record who ran each query instead of the shared account. A is incorrect because TDE protects data at rest and decrypts it for any authenticated session, so it adds no user attribution. B is incorrect because a database firewall sees the same svc_hrweb sessions, so it cannot attribute queries to individual users either. C is incorrect because rotation protects the credential but every query is still made as the same shared account.

Sample Question 3 — Fundamental Security Architecture Concepts

A threat intelligence analyst is documenting an intrusion using the Diamond Model. The notes include the command-and-control domain the implant called and the IP address of the server hosting that domain. In which vertex do these two observables belong?

  1. A. Adversary
  2. B. Capability
  3. C. Infrastructure (Correct answer)
  4. D. Victim

Correct answer: C

Explanation: C is correct because the infrastructure vertex holds the physical and logical communication structures the adversary uses to deliver and control capabilities, such as C2 domains and IP addresses. A is incorrect because the adversary vertex identifies the actor or operator behind the event, not the systems the actor uses. B is incorrect because the capability vertex describes the adversary's tools and techniques, such as the implant itself, not the servers it talks to. D is incorrect because the victim vertex describes the targeted organization, person, or asset.

Sample Question 4 — IPv6 and Modern Network Risks

An IPv6 packet capture on a user VLAN shows this message from an unknown laptop: ``` Src: fe80::a00:27ff:fe4e:66a1 Dst: ff02::1 ICMPv6 Type: 134 Hop limit: 255 Prefix: 2001:db8:66::/64 Router lifetime: 1800 ``` What is the laptop sending?

  1. A. A Router Advertisement to all nodes (Correct answer)
  2. B. A Router Solicitation to all nodes
  3. C. A Neighbor Advertisement announcing its link-layer address
  4. D. A Neighbor Solicitation for address resolution

Correct answer: A

Explanation: A is correct because ICMPv6 type 134 is a Router Advertisement, and ff02::1 is the all-nodes multicast address, so hosts may take the laptop as a default router and use its prefix. B is incorrect because Router Solicitations are type 133 and are sent by hosts to the all-routers address ff02::2. C is incorrect because Neighbor Advertisements are type 136 and do not carry prefix or router lifetime information. D is incorrect because Neighbor Solicitations are type 135 and go to a solicited-node multicast address.

Sample Question 5 — Layer 1, Layer 2, and Layer 3 Defense

An access switch begins flooding unicast traffic out of every port in VLAN 20. An engineer captures this output: ``` SW1# show mac address-table count vlan 20 Mac Entries for Vlan 20: --------------------------- Dynamic Address Count : 8187 Static Address Count : 0 Total Mac Addresses : 8187 Total Mac Address Space Available: 5 ``` Nearly all dynamic entries were learned on Gi1/0/14, a single user port. Which control most directly prevents this condition?

  1. A. Port security limiting MAC addresses per port (Correct answer)
  2. B. DHCP snooping with the uplink ports set as trusted
  3. C. Dynamic ARP Inspection enabled on VLAN 20
  4. D. BPDU Guard enabled on every access port

Correct answer: A

Explanation: A is correct because limiting how many source MACs a port can learn stops one port from filling the CAM table and forcing the switch to flood unknown unicast frames. B is incorrect because DHCP snooping blocks rogue DHCP servers and builds bindings, but it does not limit source MAC learning. C is incorrect because DAI validates ARP packets against DHCP bindings and does not stop frames with random source MACs from filling the table. D is incorrect because BPDU Guard disables a port that receives spanning tree BPDUs, which a MAC flooding tool does not need to send.

Sample Question 6 — Network Defenses, Proxies, Firewalls, and Remote Access

A network security monitoring sensor running Zeek records thousands of rows like these from one internal host within two minutes: ``` id.orig_h id.resp_h id.resp_p proto conn_state history 192.0.2.15 198.51.100.20 445 tcp S0 S 192.0.2.15 198.51.100.21 445 tcp S0 S 192.0.2.15 198.51.100.22 445 tcp S0 S ``` What does the conn_state value indicate for each row?

  1. A. The connection was established and not yet closed
  2. B. Mid-stream traffic was seen without any SYN
  3. C. The connection attempt was rejected with a reset
  4. D. A connection attempt was seen with no reply (Correct answer)

Correct answer: D

Explanation: D is correct because S0 means Zeek saw the originator's SYN but no response from the destination, consistent with scanning addresses where nothing is listening or traffic is filtered. A is incorrect because an established connection with no termination seen is recorded as S1. B is incorrect because mid-stream traffic without a SYN is recorded as OTH, and this history shows the SYN. C is incorrect because a rejected attempt is recorded as REJ, and the history would include an r for the responder's RST.

Sample Question 7 — Zero Trust Architecture and Networking

A Zero Trust project is being designed around NIST SP 800-207. A reviewer asks which logical component actually enables, monitors, and terminates the connection between a subject and an enterprise resource. Which component is it?

  1. A. Continuous diagnostics and mitigation system
  2. B. Policy Engine
  3. C. Policy Administrator
  4. D. Policy Enforcement Point (Correct answer)

Correct answer: D

Explanation: D is correct because NIST SP 800-207 defines the PEP as the component that enables, monitors, and eventually terminates connections between a subject and a resource. A is incorrect because CDM is a data source that informs the policy engine about asset state. B is incorrect because the PE makes and logs the grant or deny decision using policy and inputs, but it does not sit in the data path. C is incorrect because the PA establishes or shuts down the communication path by commanding the PEP, but it is not the component in the connection path.

Sample Question 8 — Zero Trust Endpoints and Host Hardening

A Linux server's auditd log contains these records after a sensitive file was read: ``` type=SYSCALL msg=audit(1793369112.481:2215): arch=c000003e syscall=257 success=yes exit=3 ppid=4120 pid=4188 auid=1001 uid=0 gid=0 euid=0 suid=0 fsuid=0 egid=0 tty=pts1 ses=7 comm="cat" exe="/usr/bin/cat" key="shadow_read" type=PATH msg=audit(1793369112.481:2215): item=0 name="/etc/shadow" inode=1572983 mode=0100640 ouid=0 ogid=42 nametype=NORMAL ``` The analyst connected through SSH as a normal user and then used sudo. Which field identifies that analyst's original login account?

  1. A. uid
  2. B. ses
  3. C. euid
  4. D. auid (Correct answer)

Correct answer: D

Explanation: D is correct because the audit user ID is set at login and preserved across sudo and su, so it identifies the user who originally authenticated. A is incorrect because uid is the real user ID of the process, which became 0 after sudo. B is incorrect because ses is the audit session number for the login, not a user identity. C is incorrect because euid is the effective user ID used for permission checks, which is also 0 here.

Quick 10-Question GDSA Practice Test

Take a free 10-question GDSA quick-start practice test covering all 8 GDSA domains. Get instant scoring with detailed explanations — perfect for a quick readiness check.

Why Choose FlashGenius for GDSA?

Start your free GDSA practice test now | 10-question quick start | All Sample Tests