Certified DevSecOps Engineer (ECDE) — Free Practice Questions
10 free sample questions from a bank of 10, with the correct answers and explanations. No signup required — start practising right now.
1Andrew Gerrard has recently joined an IT company located in Fairmont, California, as a DevSecOps engineer. Due to robust security and cost-effective service provided by AWS, his organization has migrated all the workloads from on-prem to AWS cloud in January of 2020. Andrew’s team leader has asked him to integrate AWS Secret Manager with Jenkins. To do so, Andrew installed the “AWS Secret Manager Credentials provider” plugin in Jenkins and configured an IAM policy in AWS that allows Jenkins to take secrets from AWS Secret manager. Which of the following file should Andrew edit to add access id and secret key parameters along with the region copied from AWS?
/etc/filebeat/filebeat.yml
/etc/sysconfig/Jenkins
/etc/file/Jenkins
/etc/sysconfig file/Jenkins
Answer: B
The short version
B - /etc/sysconfig/jenkins is the file the Jenkins Linux service sources at startup, so the AWS access key, secret key and region are injected there.
Key concepts
/etc/sysconfig/jenkins: The init-style options file shipped with the Jenkins Linux package; the service unit exports its variables before the JVM starts.
JVM system properties: -D style settings injected into the Jenkins process that the AWS Secrets Manager credentials provider plugin reads for its configuration.
AWS Secrets Manager credentials provider plugin: A Jenkins plugin that resolves credentials from AWS Secrets Manager rather than storing long-lived keys in the job or the UI.
Least privilege IAM policy: The AWS-side policy granting the Jenkins role permission to read only the specific secrets it needs, matching the plugin's credentials.
Packaging versus runtime configuration: Where a file's contents end up decides its role - package manager and log shipper configs never feed the Jenkins JVM.
Why B is correct
The plugin needs the AWS access key, secret key and region to exist in the environment of the Jenkins JVM before and during job execution. On a packaged Linux install the systemd unit sources /etc/sysconfig/jenkins and exports everything defined there into the process environment, so parameters such as AWS_ACCESS_KEY_ID, AWS_SECRET_ACCESS_KEY and AWS_REGION (or the equivalent -D properties) added to that file are available on every service start and pair with the IAM policy already in place. Of the four paths, this is the only one that is both a real Jenkins file and the correct place for service-level environment settings.
Why the others are wrong
A. /etc/filebeat/filebeat.yml belongs to Filebeat, the log shipping agent. Filebeat ships logs to an output such as Elasticsearch or Logstash; it has nothing to do with Jenkins and cannot inject credentials into the Jenkins JVM.
C. /etc/file/Jenkins does not exist on any mainstream Linux distribution. The Jenkins Debian and RPM packages install their options file under /etc/sysconfig (or /etc/default on Debian), never in a directory named /etc/file.
D. '/etc/sysconfig file/Jenkins' mangles the real path by inserting a space and turning the sysconfig directory into part of the filename. Shell paths do not work that way, so no such file can be edited.
ECDE exam tip
Two-step rule for Jenkins config questions: identify the process that must see the value, then place it in the file that process's service unit actually sources.
2Gabriel Bateman has been working as a DevSecOps engineer in an IT company that develops virtual classroom software for online teaching. He would like to clone the BDD security framework on his local machine using the following URL, https://github.com/continuumsecurity/bdd-security.git. Which of the following command should Gabriel use to clone the BBD security framework?
A - git clone https://github.com/continuumsecurity/bdd-security.git uses the correct subcommand and an exact, unaltered repository URL.
Key concepts
git clone: The git subcommand that copies a remote repository plus its history into a new local directory, here bdd-security.
Remote URL: The full path identifying the repository, including the owner - a single wrong character yields repository not found.
Subcommand structure: git is the binary and clone is the verb that follows it, separated from the URL by a space.
GitHub CLI (gh): A separate tool offering gh repo clone for authenticated GitHub operations; there is no git subcommand named github.
Exit status: Git reports failures such as an unknown remote as a fatal error and non-zero exit code, which is how a typo is detected.
Why A is correct
Cloning a remote repository means running the git binary with the clone verb and the repository's URL. Both halves of that line must be exact: 'git clone' is the recognized invocation, and the owner in the URL must be continuumsecurity, matching the URL given in the stem. Run correctly it downloads the repository and its history into a local directory named after the repository, bdd-security, with the original URL recorded as the origin remote for later fetches.
Why the others are wrong
B. The owner is misspelled as continumsecurity - the 'u' is missing from continuum. That repository does not exist, so git fails with a repository not found error even though the subcommand and the .git suffix are fine.
C. 'github' is not a git subcommand. Git resolves subcommands against its own built-ins and fails immediately with 'git: github is not a git command', so nothing is cloned. GitHub tooling lives in the separate gh CLI.
D. The same nonexistent 'github' subcommand appears here, so the command fails regardless of the otherwise correct URL spelling.
ECDE exam tip
Read a git line in three chunks - binary, verb, URL - and verify the URL character by character. Owner typos are the single most common clone failure.
3William Edwards is working as a DevSecOps engineer at SVR Software Solution Pvt. Ltd. His organization develops software products and applications related to digital marketing. William integrated Prisma Cloud with Jenkins to detect threat-intelligence based threat detection. This integration will allow him to scan container images and serverless functions for security issues in the CI/CD pipeline. Which of the following is employed by Prisma Cloud to understand the normal network behavior of each customer’s cloud environment to detect network anomalies and zero-day attacks effectively with minimal false positives?
Advanced unsupervised machine learning
Advanced supervised data mining
Advanced supervised machine learning
Advanced unsupervised data mining
Answer: A
The short version
A - Advanced unsupervised machine learning builds a behavioral baseline of each customer's own cloud environment and alerts on deviations from it.
Key concepts
Supervised learning: Requires a labeled target, so it can only detect attack classes someone has already labelled - useless for previously unseen zero-day activity.
Unsupervised learning: Derives structure from unlabeled data, allowing a model of normal behaviour to be built where no labels exist.
Behavioral / baseline profiling: Learning the normal network and workload patterns of one specific environment so its own traffic defines normal.
Anomaly detection: Scoring each observation against that baseline and flagging low-scoring deviations as suspicious.
Threat intelligence: Curated external indicators of compromise combined with the environment-specific baseline, which is what lets new attacks surface.
Why A is correct
Nobody has hand-labelled attacks inside a customer's cloud environment, and the attackers are using techniques that have never been seen before, so a supervised classifier cannot be trained for this job. Unsupervised machine learning instead fits a model of normal structure from the customer's own traffic - flows, process behaviour, connection patterns - and scores every new event against that baseline. Because the reference point is that specific environment's own behaviour, anything resembling normal activity is not flagged, which is exactly how a signature-free tool keeps false positives low while still catching zero-day anomalies.
Why the others are wrong
B. Supervised data mining requires labelled examples of attacks to learn from, and there is no such label set for zero-day activity in a customer environment. It also names the wrong technique family for behavior-based profiling.
C. Advanced supervised machine learning has the same fatal dependency: a classifier can only recognize attack patterns present in its training labels, so it cannot detect a novel zero-day that nobody has labelled yet.
D. Unsupervised is the right learning paradigm, but 'data mining' names rule and pattern discovery over stored data rather than the machine-learning anomaly scoring Prisma Cloud performs, so the technique family is wrong even though the supervision level is right.
ECDE exam tip
Unlabeled attacks means unsupervised. If a technique needs to have been taught what an attack looks like, it cannot be your zero-day defense.
4Terry Crews has been working as a DevSecOps engineer at an IT company that develops software products and web applications related to IoT devices. She integrated Sqreen RASP tool with Slack for sending notifications related to security issues to her team. How can Sqreen send notification alerts to Slack?
By creating a cookbook, defining a trigger, Alert a response, and notification
By creating a playbook, defining a trigger, security response, and notification
By creating a cookbook, defining a trigger, security response, and notification
By creating a playbook, defining a trigger, Alert a response, and notification
Answer: B
The short version
C - Sqreen builds an automation object called a cookbook, defined by a trigger, a security response and a notification.
Key concepts
Cookbook: Sqreen's automation recipe object that bundles detection, response and notification into one reusable workflow.
Trigger: The detection event that fires the recipe, such as a RASP block or an application security finding.
Security response: The action the recipe performs when triggered, such as raising an alert, blocking a request or logging the event.
Notification: The delivery step that routes the result to a third-party integration such as a Slack channel.
Integration / app connection: The authorization link to the external service (Slack, Jira, PagerDuty) that the notification step posts through.
Why C is correct
Sqreen's automation model is deliberately a three-stage recipe. A trigger watches for a security event, the security response is the action taken on that event, and the notification step pushes the outcome out to the connected integration. With the Slack app authorized on the account, the notification stage of a cookbook posts to the chosen Slack channel, so 'trigger, security response, notification' is the exact element list that makes alerts appear in Slack.
Why the others are wrong
A. This uses 'Alert a response' as the middle element, which is not part of the recipe structure. The response step in a Sqreen cookbook is the security response, not an 'alert' response.
B. The element list here is correct, but the automation object is called a playbook rather than a cookbook. Sqreen names the recipe object a cookbook, so the wrong object name invalidates the answer.
D. Two terms are wrong at once: the object is called a playbook instead of a cookbook, and the middle element is 'Alert a response' rather than the security response.
ECDE exam tip
Learn the three elements as a sentence: a cookbook has a TRIGGER, a SECURITY RESPONSE and a NOTIFICATION - anything that swaps a noun in that triad is wrong.
5Richard Branson has been working as a DevSecOps engineer in an IT company that develops apps for Android mobiles. To manage the secret information of an application in various phases of development lifecycle and to provide fine-grained access to each secret, he would like to integrate HashiCorp Vault with Jenkins. To access the vault from Jenkins, Richard installed hashicorp-vault-plugin and ran a vault instance; he then selected the AppRole authentication method, which allows apps to access vault with a predefined role. Which of the following commands should Richard use to enable AppRole authentication?
enable auth vault approle
vault auth enable approle
enable vault auth approle
auth vault enable approle
Answer: B
The short version
B - vault auth enable approle enables the AppRole authentication method at the approle path, using Vault's category-verb structure.
Key concepts
Vault CLI structure: Every command is vault, then a category (auth, secrets, kv), then an action (enable, list, login), then the path.
Auth methods: Pluggable authentication backends enabled at a path; approle is the machine-identity method built into Vault.
AppRole: Role-based authentication where an application presents a role_id and a short-lived secret_id instead of a human username and password.
Policies: Vault policies attached to a role define exactly which secret paths that AppRole role may read or write.
Jenkins plugin configuration: The hashicorp-vault-plugin reads the AppRole credentials from the job configuration once the backend is enabled.
Why B is correct
Vault's command grammar is strictly vault + category + action + path. Enabling an authentication method therefore puts auth first, the verb enable second, and the method path last - so 'vault auth enable approle' registers the AppRole backend at the approle mount and makes it available to the Jenkins plugin. Once enabled, the machine identity can log in with the role_id and secret_id pair that its bound policy authorizes.
Why the others are wrong
A. 'enable auth vault approle' scrambles the order and inserts an extra 'vault'. Vault has no top-level 'enable' command; the verb must follow the category, so this fails with an unknown command.
C. 'enable vault auth approle' has the same two defects - it starts with the verb instead of the vault binary and repeats 'vault' in the middle, neither of which matches the vault category-verb grammar.
D. 'auth vault enable approle' swaps the first two tokens. The vault binary must come first and 'auth' must be the category immediately after it, never before the binary.
ECDE exam tip
Vault verbs follow one fixed pattern - vault CATEGORY ACTION PATH - so if the line does not read that way, it will not run.
6Rachel Maddow has been working at RuizSoft Solution Pvt. Ltd. for the past 7 years as a senior DevSecOps engineer. To develop software products quickly and securely, her organization has been using AWS DevOps services. On January 1, 2022, the software development team of her organization developed a spring boot application with microservices and deployed it in AWS EC2 instance. Which of the following AWS services should Rachel use to scan the AWS workloads in EC2 instance for security issues and unintended network exposures?
AWS WAF
Amazon CloudWatch
AWS Inspector
AWS Config
Answer: C
The short version
C - AWS Inspector assesses running EC2 instances for known vulnerabilities and analyzes network reachability to surface unintended exposure.
Key concepts
AWS Inspector: A vulnerability management service that runs assessment targets against compute and reports CVEs plus network reachability findings.
Host assessment: Scanning the instance's OS, packages and installed software against the vulnerability database and reporting exploitable CVEs.
Network assessment / reachability analysis: Mapping which components on the instance are actually reachable from outside, exposing unintended attack surface.
AWS WAF: A layer 7 web application firewall filtering HTTP requests at ALB, API Gateway, CloudFront and AppSync - not a host scanner.
Amazon CloudWatch: Metrics, logs and alarms service - the telemetry layer, with no vulnerability knowledge of its own.
AWS Config: Records and evaluates configuration state against rules over time, reporting drift rather than exploitable flaws.
Why C is correct
Inspector is the AWS service built specifically for assessing workloads that are already running. It creates an assessment target against the EC2 instance, enumerates the OS and package versions, matches them to the CVE database to produce a prioritized list of exploitable vulnerabilities, and separately performs reachability analysis to show which paths into the instance are unintended. That covers both halves of the requirement - security issues and unintended network exposures - on EC2 compute.
Why the others are wrong
A. AWS WAF inspects and filters inbound HTTP requests at the edge of an application load balancer, API Gateway or CloudFront. A Spring Boot application running directly on EC2 has no WAF in front of it, and WAF knows nothing about the instance's OS packages or reachability.
B. CloudWatch collects and stores metrics, logs and alarms so you can see what is happening, and it is the right place for detection rules and notification channels - but it holds no vulnerability database and cannot scan an instance.
D. AWS Config records configuration items over time and evaluates them against managed or custom rules, so it tells you a setting has drifted from an approved baseline. It does not discover unpatched CVEs or unintended reachability on a live instance.
ECDE exam tip
Match the verb to the service: inspect/scan a running instance means Inspector, filter web traffic means WAF, watch metrics means CloudWatch, audit settings over time means Config.
7GainInsights is an IT company that develops mobile applications software. On February 11, 2022, the organization became a victim of a cyber-attack. The attacker targeted the organization’s application and compromised some important functionality. After the incident, the DevSecOps team of GainInsights identified the cause of the security issue, resolved it, and noted it for future reference. Based on this information, which of the following set of tests was conducted by GainInsights?
Security smoke tests
Security acceptance tests
White box testing
Blameless post-mortem
Answer: D
The short version
D - A blameless post-mortem: after the attack the team did root cause analysis, fixed it, and recorded the lessons without pointing fingers at individuals.
Key concepts
Blameless post-mortem: A structured incident review conducted under an explicit no-blame policy so people report the real causes rather than hiding them.
Root cause analysis: Working back from the symptom through contributing technical and process factors to the underlying cause.
Corrective actions: The fixes and safeguards derived from the review, tracked to completion so the same failure cannot recur.
Incident response lifecycle: Preparation, detection and analysis, containment, eradication and recovery, and finally the post-incident review.
Security smoke tests: Fast, shallow checks executed before a release to catch obvious breakage in a build.
White box testing: A technique that tests internal code structure and logic with knowledge of the implementation, unrelated to incident review.
Why D is correct
The stem describes the post-incident phase of incident response: an attack happened, the DevSecOps team identified the cause, remediated it, and documented it for future reference. That is precisely a post-mortem, and the 'blameless' qualifier is what makes the documentation trustworthy - without fear of punishment, people describe what really happened rather than a sanitized version, which is the only way the root cause and the corrective actions are correct. It also fits the DevSecOps culture the scenario emphasizes, where the output of a breach is a durable improvement to the pipeline.
Why the others are wrong
A. Security smoke tests are quick, shallow checks run against a candidate build before release to catch obvious breakage. They are preventive and pre-deployment, so they cannot be the activity performed after an attack has already been remediated.
B. Security acceptance tests verify that a completed build meets its stated security requirements before it goes live, and they are executed against the product rather than against an incident.
C. White box testing is a technique that exercises internal code paths with knowledge of the implementation. It is a testing strategy, not an incident review process, and nothing in the scenario describes testing code.
ECDE exam tip
An attack happened, then someone found the cause, fixed it and wrote it down - that is a post-incident review. Anything prefixed 'test' is preventive work that runs before release.
8Kevin Williamson is working as a DevSecOps engineer in an IT company located in Los Angles, California. His team has integrated Jira with Jenkins to view every issue on Jira, including the status of the latest build or successful deployment of the work to an environment. Which of the following can Kevin use to search issues on Jira?
Java query language
Jira query language
Structured query language
Atlassian query language
Answer: B
The short version
B - JQL, the JIRA Query Language, is what searches Jira issues, for example project = ABC AND status = "In Progress".
Key concepts
JQL (JIRA Query Language): Jira's declarative filter language built from field, operator and value terms joined by AND/OR, plus ORDER BY.
JIRA fields: The searchable attributes such as project, issue type, status, assignee, reporter, labels and updated.
Issue viewer integration: The Jenkins/Jira plugin view that lists issues matching a filter and annotates them with the latest build or deployment status.
REST API: Jira's programmatic interface, which accepts the same JQL string in a search endpoint, making filters reusable across plugins and dashboards.
SQL: A relational query language; Jira's backing database is internal and not exposed for administrators or users to query.
Why B is correct
Jira exposes exactly one user-facing filter language for issues: JQL. A query such as project = ABC AND status = "In Progress" AND assignee = currentUser() is parsed by the issue search service, matched against indexed issue fields, and returns matching issues - which is what the integration uses to display Jira issues alongside the status of the latest build or deployment. The same JQL string can be embedded in the plugin configuration, in dashboards and in REST API calls, which is why it is the natural answer to 'how do I search issues'.
Why the others are wrong
A. Java is the implementation language in which much of Jira is written; it is not a query interface exposed to users. There is no such thing as a 'Java query language' for searching issues.
C. SQL queries tables in a relational database. Jira's backing schema is an internal implementation detail and is not queryable by administrators or users, so SQL is not the supported way to search issues.
D. 'Atlassian query language' is not a product feature name. The language is called JIRA Query Language and is scoped to JIRA issues rather than to all Atlassian products.
ECDE exam tip
Tool-name memory hook: the language is named after the product - JIRA Query Language - and it is a filter over issue fields, not a database query.
9William Friedkin has been working as a DevSecOps engineer in an IT company for the past 3 years. His team leader has asked him to validate the host configuration that runs the Docker containers and perform security checks at the container level by implementing Docker’s CIS Benchmark Recommendations. Therefore, William would like to integrate Docker Bench with Jenkins to incorporate security testing in DevOps workflow and secure the Docker Container. Before starting the procedure, he would like to install openssh on Ubuntu. Which of the following command should William run to install openssh on Ubuntu?
sudo apt-get -s install openssh-server
sudo apt-get install openssh-server
sudo apt.get install openssh-server
sudo apt.get -s install openssh-server
Answer: B
The short version
B - sudo apt-get install openssh-server performs the real installation; the -s flag only simulates the transaction.
Key concepts
apt-get: The command-line package manager for Debian and Ubuntu, invoked as apt-get with an action such as install, update or remove.
apt versus apt-get: apt is the human-friendly front end whose command-line options may change between releases; apt-get is the stable, scriptable interface.
-s / --simulate: The dry-run flag that resolves dependencies and prints what would happen without downloading or installing anything.
sudo: Runs the command with root privileges, required because package installation writes to system directories.
Package name: openssh-server is the Ubuntu package providing the SSH daemon; the client side is the separate openssh-client package.
Why B is correct
On Ubuntu the package manager is apt-get, and its syntax is the binary followed by an action and the package name, executed with root privileges via sudo. So 'sudo apt-get install openssh-server' resolves the package and its dependencies, downloads the .deb files and configures the SSH daemon. Adding -s turns that entire sequence into a simulation that only prints the plan, which is exactly the trap option A sets - and apt.get with a dot is not a real command at all.
Why the others are wrong
A. The -s flag makes apt-get perform a simulated run: it resolves dependencies and prints what it would do, but nothing is downloaded or installed. The server package is never actually placed on the host.
C. The binary is named apt-get, not apt.get. A dot in the command name makes the shell report command not found, so no installation is attempted whatsoever.
D. This option stacks both errors: the nonexistent apt.get binary and the -s simulation flag. It fails before it can do anything.
ECDE exam tip
Any command that contains 'apt.' or ends in '-s' is a decoy. The real installer is always 'sudo apt-get install <package>'.
10William O’Neil has been working as a senior DevSecOps engineer in an IT company that develops software products related to ecommerce. At this point in time, his team is working on securing a python-based application. Using GitGraber, William would like to detect sensitive information in real-time in his organizational GitHub repository. Therefore, he downloaded GitGraber and installed the dependencies. Which of the following commands should William use to find secrets using a keyword (assume the keyword is yahoo)?
D - The -w flag points GitGraber at a wordlist of keywords, -q supplies the keyword pattern, and -s runs the repository scan.
Key concepts
-w / --wordlist: Takes a file containing a list of keywords or strings to search for, one per line.
-q / --query: Takes the keyword or regular expression pattern being hunted for, here yahoo.
-s / --scan: Runs the secret scan over the supplied Git repository and reports matches, which is what enables real-time detection in CI.
-k / --keywords: Supplies keywords directly as command-line arguments rather than from a file.
-p / --patterns: Takes a file of custom regular expressions describing secret formats to look for.
-g / --git-token: Supplies GitHub personal access token(s), optionally from a token file, for authenticated scanning of private repositories.
Why D is correct
The requirement is to search by keyword taken from a file. GitGraber distinguishes that from several similar options: -w consumes a wordlist file, which is exactly what wordlist/keywordsfile.txt is, while -q supplies the specific keyword pattern and -s triggers the scan of the repository so matches are reported as they are found. Running that on the organization's repository inside CI gives the real-time secret detection the scenario asks for.
Why the others are wrong
A. -p reads a patterns file, which holds custom regular expressions that describe secret formats such as provider-specific token shapes. Feeding it a plain keyword list does not perform a keyword search, so the yahoo keyword would not be matched as intended.
B. -g is for GitHub personal access tokens, optionally loaded from a token file, and authenticates the scan against the API. A keyword wordlist is not a token list, so this both fails to search and misuses the credential option.
C. -k expects keywords as arguments on the command line. Passing the path wordlist/keywordsfile.txt there makes GitGraber search for that literal string as a keyword rather than reading the list inside the file.
ECDE exam tip
Flag-to-input memory hook: -w wordlist file, -k keywords on the command line, -p pattern file, -g git token, -q the query itself, -s scan.