Concepts

Network Security Foundations

The vocabulary and mental models every later topic builds on: what we protect, what threatens it, how weaknesses become risk, and why layered, minimal, deliberate design beats any single control.

01What network security is

Network security is the combination of controls that keep networked systems and the data moving through them trustworthy. It is not a product or a perimeter; it is a continuous loop. Some controls stop bad things from happening, others notice when they happen anyway, and still others limit the damage and restore service.

Prevent Detect Respond Recover lessons learned feed the next cycle firewalls, patching,authentication backups, failover,rebuild from known-good
Security as a loop. The same outcome functions appear in the NIST Cybersecurity Framework (Govern, Identify, Protect, Detect, Respond, Recover).

Every control lives in one of three domains, and attackers probe all three. A flawless firewall is irrelevant if an employee reads a password over the phone to a stranger.

People

Awareness, incentives, habits, social engineering resistance.

Process

Change control, access reviews, patch cadence, incident playbooks.

Technology

Encryption, filtering, monitoring, identity systems, hardened configs.

02The CIA triad

Three properties define what "secure" means for data and services. Almost every attack can be described as a violation of one or more of them, and almost every control exists to protect one or more of them.

Confidentialitywho can read Integritywho can change Availabilitywho can use, when Protected asset data · service · device encryption, access control redundancy, DDoS mitigation hashes, signatures, MACs, change control
Controls sit on the edges: many protect two properties at once, and some trade one against another.

The properties compete. Encrypting and locking everything down strengthens confidentiality but adds failure points that hurt availability. The right balance depends on what the asset is for.

AssetDominant propertyWhy
Password vaultCDisclosure is catastrophic and irreversible.
Software update serverIA tampered update is trusted and installed everywhere.
Grid control networkADowntime has physical, real-time consequences.
Public marketing siteA IContent is public; defacement and outages are the harm.

Two companion properties are often added: authenticity (the party or data is genuinely who or what it claims) and non-repudiation (an actor cannot credibly deny an action). Both matter heavily in network protocols.

03Threat, vulnerability, risk

These words are used loosely in the press but mean different things, and mixing them up leads to spending on the wrong problems.

Threat

Something that could cause harm: an actor, an event, or a technique. Exists whether or not you are weak.

Vulnerability

A weakness a threat could use: a bug, a misconfiguration, a missing process, a gullible habit.

Asset

What you stand to lose: data, uptime, reputation, safety, money.

Risk

The expected harm when a threat meets a vulnerability on an asset you care about.

Threat Vulnerability Asset Risk Risk ≈ likelihood × impact remove any circle and the overlap disappears
You rarely control threats. You can shrink vulnerabilities and reduce what an asset exposes, which is why most defensive work targets those two circles.

Once risk is estimated, there are only four responses: mitigate it with controls, transfer it (insurance, contracts), avoid it by not doing the risky thing, or accept it knowingly.

04The threat landscape

The threat landscape is the current set of adversaries, techniques, and exposures facing a system. It shifts constantly as technology changes (cloud, remote work, IoT), as criminal business models mature (ransomware-as-a-service), and as geopolitics changes who is targeted.

ExternalInternal MaliciousAccidental Nation-state Ransomware crew Hacktivist Rogue insider Careless admin Vendor misconfig Fat-finger deletion
Classify threats by origin and intent, then by method (technical vs. human). Accidental insiders cause a large share of real incidents.

Major categories

Malware

Worms, trojans, ransomware, info-stealers. Self-spreading variants turn one host into many.

Social engineering

Phishing, pretexting, MFA fatigue. Attacks the person, not the protocol.

Network attacks

Scanning, spoofing, interception, denial of service against exposed services.

Supply chain

Compromise a trusted vendor or update channel and inherit its access.

Two illustrative incidents

Supply chain: in 2020, attackers inserted a backdoor into signed updates of a widely used IT monitoring product, reaching thousands of organizations through a channel they had every reason to trust. The core failure was integrity of the build pipeline.

Credential abuse: in 2021, a single reused password on a legacy remote-access account without MFA opened the door to ransomware that halted a major US fuel pipeline. A simple lapse produced an availability crisis with national effects.

05Attack surface

The attack surface is every point where an attacker can send input to, or pull data from, a system. Each open port, running service, API endpoint, login form, admin account, and reachable person is a door. More doors means more to defend and more chances one is left unlocked.

Before After reduction Server telnet, ftp, smb, rdp, default admin, debug API, test pages… Server https (public), ssh (VPN-only, key auth)
Reduction: disable unused services, close ports, remove default accounts, validate every input, and keep internal systems off the public internet.

A useful exercise is enumerating your own surface the way an attacker would, with a port scanner against a lab host you own.

06Defense in depth

Every control fails sometimes: signatures miss new malware, users click, patches lag. Defense in depth assumes this and stacks independent layers so that an attacker has to beat all of them in sequence, while defenders get multiple chances to notice.

blocked perimeter segmentation authentication endpoint monitoring attack path
Each layer has holes. Breach requires the holes to line up; independent layers make that unlikely.

The layers should be diverse, not duplicates. Two firewalls from the same vendor with the same rule set are closer to one layer than two.

Modern networks push this further with zero trust: no request is trusted because of where it comes from on the network; every access is authenticated, authorized, and evaluated in context.

07Least privilege

Every user, service, and process should hold exactly the permissions its job needs, for only as long as it needs them. When something is compromised or makes a mistake, the damage is capped by what it was allowed to do.

Broad access Least privilege web app HR db finance backups AD web app HR db finance backups AD shaded circle = blast radius if the web app is compromised
Same breach, very different outcome.

In practice this means role-based access, separate admin accounts, services running as unprivileged users, scoped API tokens, just-in-time elevation, and regular reviews that remove permissions nobody uses. It pairs with separation of duties: no single identity can both request and approve a sensitive action.

08Threat modeling

Threat modeling is structured imagination: before building or defending a system, ask what it does, what could go wrong, what you will do about it, and whether you did a good job. Two tools make that repeatable.

STRIDE

A checklist of six threat types, each the inverse of a security property. Walk each component and data flow in a diagram and ask which of the six apply.

ThreatViolatesNetwork example
SpoofingauthenticityForged source IP or ARP reply
TamperingintegrityAltering packets in transit
Repudiationnon-repudiationAdmin action with no audit log
Information disclosureconfidentialitySniffing cleartext credentials
Denial of serviceavailabilitySYN flood exhausting a server
Elevation of privilegeauthorizationExploiting a service running as root

Attack trees

The attacker's goal is the root; children are ways to achieve it. OR nodes need any one child, AND nodes need all. Annotating leaves with cost or difficulty reveals the cheapest path, which is where defense should go first.

Read a user's email OR Steal password Intercept traffic Compromise server ORAND phish · cheap brute force get on path defeat TLS unpatched CVE cheapest leaf (highlighted) tells you to prioritize MFA and phishing resistance
AND branches are expensive for attackers; a single cheap OR leaf is usually the real risk.

09Naming and scoring vulnerabilities

Thousands of vulnerabilities are disclosed every year. A shared vocabulary lets scanners, vendors, and defenders talk about the same flaw and decide what to fix first.

CWEweakness type CVEspecific instance CVSSseverity 0–10 KEV / EPSSexploited? likely? e.g. buffer overflow CVE-YYYY-NNNNN how bad in theory how urgent in practice
Severity is not priority. A medium-scored flaw being actively exploited on an internet-facing host outranks a critical one on an isolated lab box.

Vulnerability scanners automate the first pass: they fingerprint services and match versions against these databases. Their output is a starting point for analysis, not a verdict; false positives and context-blind scores are common.

10No perfect security

Security is risk management under constraints, not a finished state. Every control costs money, time, and convenience, and controls people find unbearable get bypassed.

Security Usability Low cost chosen operating point moving toward one corner pulls away from the others
The design question is where to sit for this asset and this threat, not how to reach a corner.

Common misconceptions

MythSecurity is a technical problem.
RealityMost breaches involve a human step: a click, a reused password, a skipped process.
MythA firewall and antivirus are enough.
RealityNo single control covers novel exploits, encrypted payloads, or stolen credentials.
MythWe're too small to be a target.
RealityAutomated scanning is indiscriminate, and small firms are stepping stones into larger partners.
MythStricter is always safer.
RealityFriction breeds workarounds; security designed around real workflows holds up better.

—References

Peer-reviewed

  1. Adams, A. & Sasse, M. A. (1999). Users are not the enemy. CACM, 42(12), 40–46. doi:10.1145/322796.322806
  2. Anderson, R. (2001). Why information security is hard — an economic perspective. ACSAC. doi:10.1109/ACSAC.2001.991552
  3. Bilge, L. & Dumitraş, T. (2012). Before we knew it. ACM CCS. doi:10.1145/2382196.2382284
  4. Herley, C. (2009). So long, and no thanks for the externalities. NSPW. doi:10.1145/1719030.1719050
  5. Jacobs, J. et al. (2021). Exploit Prediction Scoring System (EPSS). DTRAP, 2(3). doi:10.1145/3436242
  6. Kordy, B., Piètre-Cambacédès, L. & Schweitzer, P. (2014). DAG-based attack and defense modeling. Computer Science Review, 13–14. doi:10.1016/j.cosrev.2014.07.001
  7. Manadhata, P. & Wing, J. (2011). An attack surface metric. IEEE TSE, 37(3). doi:10.1109/TSE.2010.60
  8. Mauw, S. & Oostdijk, M. (2006). Foundations of attack trees. ICISC 2005. doi:10.1007/11734727_17
  9. Ohm, M., Plate, H., Sykosch, A. & Meier, M. (2020). Backstabber's knife collection. DIMVA. doi:10.1007/978-3-030-52683-2_2
  10. Reason, J. (2000). Human error: models and management. BMJ, 320, 768–770. doi:10.1136/bmj.320.7237.768
  11. Saltzer, J. & Schroeder, M. (1975). The protection of information in computer systems. Proc. IEEE, 63(9). doi:10.1109/PROC.1975.9939

Standards and frameworks