Back to blog

Security+ Flashcards: Card the Control, Not the Acronym

Most Security+ decks are acronym glossaries. The exam hands you the acronym. Here is how to card what a control does, when you would pick it, and its lookalikes.

If your Security+ deck is four hundred cards that look like "SAML: Security Assertion Markup Language", you have written a glossary and called it a deck. The exam already gives you the acronym. It hands it to you inside a paragraph about a company that wants employees to click into a cloud HR app and land already logged in, then asks which of four things you would deploy. Expanding the letters gets you no closer to answering that.

So here is the whole argument in one line: the acronym is the cue on the front of the card, not the answer on the back. The back has to say what the control does, when you would reach for it, and which control it keeps getting mistaken for. Those are three different cards, and the third one is the one almost nobody writes.

I work on Nebulearn, a flashcard app, so read the tool paragraph near the end with the suspicion it deserves. I should also say plainly that I have not sat Security+ and I am not a security professional. I am an engineering student, and the closest thing I have is a computer networks course where I spent weeks confidently mixing up protocol acronyms I could expand perfectly. That is a card design problem, not a security knowledge problem, and it is the same one that shows up in medical abbreviation decks.

The exam gives you the acronym in the question

Expansion cards are seductive because they are the easiest cards in the world to make. The source hands you a bolded term and a parenthetical, the front and back write themselves, an AI generator will produce a hundred of them off one chapter, and grading yourself is unambiguous. You either produced the words or you did not. An evening of that feels like enormous progress.

The recall is real. It is just recall of a lookup the exam performs for you for free. A question stem that says "the company wants to implement RADIUS" has already told you the letters. What it wants to know is whether you understand that you are looking at network access authentication rather than browser web single sign-on, and whether the scenario it just described is one where that is the right call.

Two honest exceptions, because a rule you cannot state the limits of is not a rule. First, some expansions genuinely are the definition. Single sign-on, mandatory access control, least privilege, separation of duties. The words do the work, and a card asking what the term means ends up asking about behaviour whether you meant it to or not. Keep one card for those and move on.

Second, you need a recognition floor. If you read TACACS+ and feel nothing at all, you cannot reason through a scenario, because you have no handle to grab. A plain expansion card is the right fix for that state. What it is not is the deck. Tag those as recognition cards and stop adding to the pile the moment the letters stop being noise. Everything past that point is a card for something you already know.

Which raises the obvious question. If one definition card is not enough, what do the others ask?

Four cards per control, and only one is a definition

Take SAML as the worked case, because it sits in the middle of the messiest acronym neighbourhood on the exam. Here is the set I would actually write.

The function card asks what it does in one sentence. Front: "What does SAML do?" Back: "An identity provider sends a signed assertion about an already-authenticated user to a service provider, so the user does not log in again at the service." That is the definition card, and it is one of four rather than the whole entry.

The choice card asks when you would pick it. Front: "Employees should click a link to a third-party web app and arrive already signed in, using the corporate directory. What are you deploying?" Back: "SAML-based web single sign-on, because the requirement is a browser handoff of an existing corporate identity to an external web application." Now the acronym is the answer instead of the question, which is the direction the exam runs.

The discrimination card asks what it gets confused with. Front: "A third-party app needs to read a user's calendar on their behalf. SAML or OAuth, and why not the other?" Back: "OAuth, because the app needs delegated authorization to an API, not a login handoff. OIDC is the layer on top of OAuth that adds the identity token, and that is the part people mistake for SAML."

The boundary card asks where the thing stops. Front: "Why is RADIUS not the answer to a web single sign-on question?" Back: "RADIUS authenticates access to the network itself: wireless, VPN, switch ports. Not a handoff between two web applications."

Four cards for one acronym. Run that across an objectives subsection holding eighteen acronyms and you are not writing eighteen cards, you are writing something closer to forty, because the exceptions above collapse some of them and one shared discrimination card covers two terms at once. That is arithmetic off my own rule, not a measured study of anything. The honest number is bigger than the glossary number and smaller than four times everything, which is exactly why the expansion deck keeps winning on effort alone.

Confusion pairs belong on one card, not two

The discrimination card is the one I want to argue for hardest, because two separate cards for two lookalike controls are actively worse than one card for the pair. When RPO gets its own card, the answer is never in doubt. The front says RPO, so the back is about RPO. You practise retrieval in a setting where nothing competes, and then the exam puts both terms in the same list of four options and the distinction you never practised is the entire question. That is context dependency doing what it always does.

So write the pair as one card with a front that forces a choice. These are the pairs I would build first.

PairWhat actually separates themCard that fixes it
RPO vs RTOData you can afford to lose vs time until service is back"Backups run nightly at 2am. Which number does that set?"
IDS vs IPSAlerts out of band vs blocks inline"The malicious packet never reaches the server. Which one, and what changed about its position?"
RBAC vs ABACJob role vs attributes of the request"A contractor gets read access, company laptop only, business hours only. Which model, and why does the other get ugly?"
RADIUS vs TACACS plusNetwork access AAA vs administrative access to devices"An engineer logs into a switch CLI and every command is logged. Which one, and why?"
Hashing vs encryptionOne way, for integrity vs reversible with a key, for confidentiality"You need to show a log file was not altered. Which, and why is the other wrong here?"
Vulnerability scan vs penetration testFinds known weaknesses vs authorized exploitation of them"The scope document authorizes exploitation. Which activity, and who signs it off?"

Look down the middle column and notice that every entry is one sentence long. That is not me compressing for a table, it is the test for whether you have a discrimination card at all. If you cannot write the separating sentence in one clause, you do not understand the distinction well enough to card it yet, and what you write instead will be two definitions sitting next to each other hoping you pick up the difference on your own. Go back to the source, find the sentence, then write the card. The same rule is what makes any card stick: the card is only as sharp as the sentence underneath it.

Four-way groups are where this gets genuinely hard, and the recovery metrics are the standard trap: RPO, RTO, MTTR, MTBF. Do not build one card with four options on the front. It will keep failing, because you are asking for four retrievals inside one grading decision. Build a chain instead. RPO against RTO first. MTTR against MTBF second, once you have the repair versus failure axis straight. Then one sorting card whose front is just "sort these four into the two that describe recovery targets and the two that describe reliability history." Three cards with one job each, instead of one card that punishes you for a blank on any of four facts.

When this breaks

It breaks in four predictable ways, and three of them are fixable without abandoning the method.

The discrimination card becomes a leech. You fail RADIUS against TACACS plus every time it comes up and you start dreading the card. Nine times out of ten that card is secretly carrying two facts, usually transport and encryption details riding along with the use case split. Break it apart: one card for the use case, one for the protocol detail if the objectives actually ask for it. The general version is the leech protocol, and the diagnosis it starts from is the right one here. A card you always fail is usually badly encoded, not badly memorized.

The stems get long. Once you start writing scenario fronts it is easy to slide into writing full exam questions, four distractors and a three-sentence company backstory. Then a single review takes ninety seconds, you get through ten, and you quit. Cap the front at two sentences and make it answerable out loud in a few seconds. That is the atomic encoding rule applied to scenarios: one decision per card, even when the decision needs a little setup.

The deck has holes you cannot see. If you built it from one video series plus your own notes, your coverage is that instructor's emphasis, not the exam's. Use the objectives PDF that CompTIA publishes for the current version as the index and check your tags against it domain by domain. Any domain with no discrimination cards in it is a gap, not a mastered section. That is the part of an exam deck audit people skip, because a deck that feels finished is very convincing.

The fourth failure is not a card problem. If you sit down to write the separating sentence for two controls and simply cannot, no format fixes that. Close the card editor and go read the section. Flashcards hold something you have already understood once, and no encoding trick substitutes for that.

Where a premade deck still wins

I would be lying if I said writing forty cards per objectives subsection is the fast path, so here is the honest alternative. Mature community Security+ decks exist for Anki, they are free, they run on your phone, and if the real choice is between downloading one tonight and writing your own next week, download the Anki deck. Quizlet wins a different case: three people splitting the domains before a Friday session, one shared set, nobody configuring anything.

The catch with an inherited deck is that its discrimination cards encode somebody else's confusions rather than yours, which is the standard problem with using someone else's deck. You still write your own cards for the pairs that trip you specifically. Review history also does not follow you between apps, so moving later restarts the schedule even when the cards come with you.

The bottleneck on this page is narrow. Writing choice and discrimination cards means keeping the source paragraph in view while you rewrite it as a decision, and most tools make you tab away from the document to do it. That is what Nebulearn's PDF upload with the side-by-side view is for: the objectives PDF stays open beside the card you are writing, so the scenario you invent still matches what the source said. When an AI draft comes back with a pile of expansion cards you do not want, skip or ignore them instead of deleting one at a time. FSRS is on by default, which matters more than usual here, because recognition cards and discrimination cards should not mature at the same rate and you should not be deciding that by hand.

Glossary decks are quick to build and they test a lookup the exam does for you. Control decks are slower to build and they test what the exam actually asks. If the second one is what you want and the writing cost is what has been stopping you, that gap is the one Nebulearn is built to close.

Security+CertificationFlashcardsStudy Tips