Country or region and language

    Choose your country or region and your language.

    Suggested for you

      All countries and regions

        Choose your character

        It shows in the top bar and on your profile. Change it any time.

        Application security

        Threat model a feature, review code for flaws and decide what gets fixed first.

        • 7 scenarios
        • 35 minutes each
        • Speak or type your answers
        • A report at your level

        A computer voice asks the questions.

        Your interview level

        Practice scenarios

        Choose a scenario. Each interview ends with a report for the level you selected.

        35 min4 stagesMid level

        Threat model a password reset flow

        Larkhill Pay is adding a self-service password reset. A customer enters an email address, receives a link and sets a new password; support staff can also start a reset from an admin screen. Threat model the flow before it ships.

        Start interview
        Rubric and preparationWhat a mid answer covers · 4 lessons to prepare

        Mid standard

        • A token that can be guessed or reused lets an attacker take over the account, and a token that leaks through logs, referrer headers or a shared inbox does the same. I would require a long random single-use token with a short expiry because a leak then has a small window, and I would invalidate it once used or when a newer one is issued.
        • I would give the same response and timing whether or not the account exists, because otherwise the form becomes a way to list customers through enumeration. I would add a rate limit per address and per source and alert on bursts, because the form sends email on demand.
        • I would require support to verify identity with a documented check, because a phone call alone can be social engineering. The action needs an audit record naming the agent, the owner should get a notify message at their existing contact, and I would test that a second reset invalidates the first link.
        35 min4 stagesMid level

        An order endpoint with broken access control

        Gable Freight exposes GET /api/orders/{orderId} to its customers. A code review shows the handler checks that the session is valid, then loads the order by its ID. Order IDs are sequential numbers.

        Start interview
        Rubric and preparationWhat a mid answer covers · 4 lessons to prepare

        Mid standard

        • I would put the check on the server, in the query or one shared authorization function, so the order loads only where its customer matches the session’s customer. I would return the same not found response either way, because a different response tells an attacker the order exists.
        • I would test with two accounts, because a single test customer can never show one customer reaching another’s data: A reads her own order, A asking for B’s order gets not found, and an unauthenticated request is rejected. I would repeat those tests for the list, update and delete endpoints.
        • I would check access logs for sessions that read many different order IDs, because that pattern shows whether the flaw was used and how much exposure there was. Then I would bring in the incident and privacy owners, since customer data may have been read, ship the fix first and leave the notification decision with its owners.
        35 min4 stagesMid level

        A report search with string-built SQL

        Copperfield Bank’s internal report tool builds its query by joining a branch name the user types into a SQL string. A developer proposes escaping single quotes to fix it. Review the risk and the fix.

        Start interview
        Rubric and preparationWhat a mid answer covers · 4 lessons to prepare

        Mid standard

        • Escaping is easy to get wrong for encoding, numbers and identifiers, because the query is still built from strings. I would use parameterized queries, which send the value separate from the SQL, so input can never become code.
        • I would apply least privilege by giving the report account read-only access to only the tables it needs, because an injection that slips through somewhere else then does less harm. I would also check the branch against an allowlist of real branches and stop returning raw database errors.
        • I would search the code for queries built by concatenation or template strings, then add a static analysis rule to the pipeline so new ones fail review. I would fix them in order of risk, starting with anything reachable from outside, because exposure decides the damage.
        35 min4 stagesMid level

        Too many dependency findings for one sprint

        Nettle Health’s dependency scanner reports 140 findings across 12 services: 6 critical, 30 high and the rest medium or low. The team can fix about 15 this sprint. Decide what gets fixed first.

        Start interview
        Rubric and preparationWhat a mid answer covers · 4 lessons to prepare

        Mid standard

        • I would trace the call path to see whether our code calls the affected function, check the configuration that turns the feature on and confirm with the service owner. If I cannot show it is unreachable, I treat it as reachable, and I record the evidence because someone will review that decision later.
        • I would apply a mitigation now, such as turning off the feature or blocking the input at the edge, because the upgrade will take weeks. I would plan the upgrade with the owning team against a date and record the exception with an owner and an expiry so it does not become permanent.
        • I would agree a fix deadline for each severity and exposure level, automate small dependency updates and block new critical findings at build time with a documented override. I would report overdue findings by team every month, because what nobody sees does not get fixed.
        35 min4 stagesMid level

        Stored script injection in teacher comments

        Tallow Learning lets teachers post comments that appear on each student’s dashboard. A tester posted a comment containing a script tag, and it ran in a student’s browser. The comment is stored and then inserted into the page as HTML.

        Start interview
        Rubric and preparationWhat a mid answer covers · 4 lessons to prepare

        Mid standard

        • I would encode the output for the context where it is shown, rendering the comment as text, not HTML, because the browser then never runs it. If teachers need formatting, I would sanitize it on the server with a maintained allowlist library, since filtering input alone misses new output contexts.
        • I would add a content security policy that blocks inline and unknown scripts, and mark the session cookie HttpOnly and Secure, because then an injected script can neither load new code nor read the cookie. These do not fix the bug, but they limit what an injected script can do.
        • I would test comments with script tags, event handler attributes and encoded variants, because attackers try more than the obvious tag, and check that each shows as text. Then I would check every other place the comment appears, such as notification email and the export, since each is a separate output context.
        35 min4 stagesMid level

        A cloud key committed to a public repository

        At Brindle Logistics, a developer notices that a cloud access key was committed to a public repository three days ago and removed in a later commit. The key belongs to a service account that can read and write shipment files.

        Start interview
        Rubric and preparationWhat a mid answer covers · 4 lessons to prepare

        Mid standard

        • I would check the access logs for that key across the whole exposure window, because they show which actions ran, from which addresses and against which files. I would compare that with the service’s normal pattern and escalate to incident response if anything does not match.
        • I would turn on secret scanning in the repository and in pre-commit hooks, and move the service to short-lived credentials from a secrets manager, because a short-lived credential is useless soon after it leaks. I would also narrow the account’s permissions to what it actually needs.
        • I would run a blameless review because people who fear blame hide the next leak: thank the developer for the report, record what let the key reach the repository and share the new guardrails. The review covers the gap in our process, not the person.
        35 min4 stagesMid level

        Session tokens for a new patient app

        Quarry Health is launching a mobile app beside its patient website. The team proposes long-lived tokens stored on the phone so patients rarely sign in again. Review the session design before launch.

        Start interview
        Rubric and preparationWhat a mid answer covers · 4 lessons to prepare

        Mid standard

        • I would use short-lived access tokens with a refresh token kept in the phone’s secure storage, because a stolen access token then expires quickly. I would rotate the refresh token on each use and keep the ability to revoke it on the server, so a reused refresh token can be detected and blocked.
        • I would keep tokens in the platform keystore, never in plain files or logs, and send them only over TLS in an authorization header, because a token in a URL ends up in logs. I would keep tokens out of analytics events and crash reports too.
        • I would test that expired and revoked tokens are rejected, that a password change ends other sessions and that a replay of a refresh token is blocked, because each of those is a way a stolen token stays useful. I would put these in the security review and the automated suite.

        Interview Hub access

        The Interview Hub is included in All Access. Choose a plan or sign in to your account to continue.

        View plansSign in or join

        Your Work Gym account

        Save your progress and pick up on another device. Your browser's progress comes with you.

        or

        You can keep practicing without an account.

        Help and feedback

        Ask a question, tell us what is not working, or share an idea. We reply by email.

        What is it about?

        We use it only to reply to you.