MODULE 2 · LESSON 1
Free — no login requiredSign in to track progress, save quiz attempts and enrol in the full course.
Sign in to track progress / enrolThe Three Properties Worth Protecting
Every security control ever built protects one of three properties. The model is normally called the CIA triad, which has nothing to do with the intelligence agency and everything to do with three words.
Confidentiality: can only the right people see it?
The property most people mean when they say security. Breaking it produces a leak.
A working example. Your HR team stores salary data on a shared drive. Confidentiality is the question of whether the permissions on that drive actually match who is supposed to see it, which in many organisations is a genuinely uncomfortable question the first time somebody checks.
The AI era version: an engineer pastes unreleased product specifications into a public chatbot to get help writing a summary. Nobody attacked anything. Confidential material simply left the building, carried out by a helpful employee. Module 6 covers this in depth, because it is now one of the most common ways organisations lose confidential data.
Integrity: is it still true and unaltered?
The property nobody thinks about until it fails, and the one that tends to cause the most expensive damage.
Integrity is about whether data can be changed by someone who should not be able to change it, and whether you would find out. Note the second half. A leak announces itself eventually. Silent corruption does not.
A working example. An attacker who reaches your finance system could steal the payment records, which is a confidentiality failure and would probably be noticed. Alternatively they could quietly alter the bank details on a single supplier record. Payments then flow to the attacker's account, correctly processed, fully authorised, appearing in every report as normal, until the real supplier complains about three months of missing money.
The AI era version is genuinely new and worth understanding. If your organisation runs an assistant that answers questions from an internal knowledge base, then anyone able to edit that knowledge base can influence what the assistant tells everyone who asks. The attack does not touch the model. It edits a page. Every subsequent answer is confidently wrong in the direction the attacker chose.
Availability: can you reach it when you need it?
The property that turns a security incident into an operational crisis.
A working example is ransomware, which is fundamentally an availability attack. The data is not stolen in the classic sense, it is still there, you simply cannot get to it. This is why backups are the definitive ransomware control, and why the only backup that counts is one that has been restored from in a test. An untested backup is a belief, not a control.
Availability failures also arrive without any attacker. A single expired certificate, a misconfigured update, or a provider outage produces the same result.
Using the model in practice
The value of the triad is that it converts a vague worry into three specific questions you can ask about anything, without needing technical depth.
Suppose a colleague proposes connecting an AI note taking assistant to every meeting recording. You do not need to understand how the assistant works to ask:
- Confidentiality. Who will be able to read these summaries and where are they stored? Meeting recordings contain salary discussions, legal matters, personnel issues and unreleased plans. This creates a single searchable store of all of it.
- Integrity. If a summary is wrong, who notices? Summaries get forwarded and quoted, and an error acquires authority as it travels.
- Availability. What happens to us if this is unavailable, or if the vendor discontinues it and we have built a process on it?
Those three questions are not obstruction. They are the conversation that should happen before adoption rather than after an incident.
Beginners often assume more security is simply better. In practice the three properties trade off against one another, and most genuine security decisions are choices about which to favour.
Confidentiality against availability. The most confidential system is one nobody can reach. Every access restriction you add makes legitimate work slightly harder. Push too far and people build workarounds, which is how shadow systems appear: the official route was so painful that the team invented an unofficial one with no controls at all. Security that is too strict frequently produces less security overall.
Integrity against availability. Requiring a second approver on payments protects integrity and slows every payment down. That is a real cost, and the right answer depends on the size of the payment. Sensible organisations set thresholds rather than applying one rule to everything.
Confidentiality against integrity. Detailed audit logs let you detect and reconstruct tampering, which serves integrity. Those same logs record who did what and when, which is itself sensitive, and a compromised log store is a rich target. You have improved integrity by creating a new confidentiality problem.
This is why security work is rarely about finding the secure option. It is about deciding which property matters most for this specific asset and accepting the cost that follows. A public marketing brochure needs integrity and availability but essentially no confidentiality. A private encryption key needs overwhelming confidentiality and almost nothing else. Treating both the same way wastes effort in one place and leaves a hole in the other.
When someone proposes a control that has no cost at all, look more carefully. Either the cost is being paid by somebody who is not in the room, or the control does very little.
An attacker gains access to a supplier database and changes the bank account details on one record, leaving everything else untouched. Which property has been broken, and why is this case particularly dangerous?
Confidentiality
Click to flipOnly authorised parties can read the information. Breaking it produces a leak, which usually becomes visible eventually.
Click to flip backEvery control protects confidentiality, integrity or availability. Confidentiality failures leak, availability failures stop work, and integrity failures do neither, which is why they run longest and cost most. Run all three as questions over anything new, including AI tools you do not understand technically: who can now see this, who can change it and would we notice, and what happens to us if it stops. The three properties also pull against each other, so the goal is never maximum security but a deliberate choice about which property this particular asset needs most.