This is Chapter 1 of a full series.
The complete roadmap (networking → Linux → AWS → DevOps → DevSecOps → Cloud Security Engineering → Architecture) lives here:
👉 AWS Cloud Security Roadmap: The Full Learning Path from Networking to Security Engineering

Roadmap on Medium, hands-on chapters here on dev.to. Follow the series so you don't miss the next one.

Why start here?

Cloud security sits on top of networking. If you can't explain what happens between typing a URL and seeing a page, you can't debug a security group, read a VPC Flow Log, or understand why a TLS certificate matters.

Goal of this chapter: you can trace one request end to end, calculate a CIDR block by hand, and diagnose DNS, port and certificate problems from the command line.

The big picture

Every layer can fail on its own, with its own symptoms. That's the whole skill: know which layer is broken.

1. DNS: names to IPs

Resolution chain: stub resolver (your OS) → recursive resolver (ISP / 1.1.1.1 / 8.8.8.8) → root servers → TLD servers (.com) → authoritative server for the domain.

Record Purpose Example
A name → IPv4 example.com → 93.184.216.34
AAAA name → IPv6 example.com → 2606:2800:...
CNAME alias to another name www → example.com
TXT free text (SPF, domain verification) "v=spf1 ..."
MX mail servers 10 mail.example.com

TTL = how many seconds resolvers may cache the answer. Low TTL = fast changes, more queries. High TTL = fewer queries, slow propagation. This is why "I changed the DNS but nothing happened" is usually just cache.

Rule: a CNAME can't coexist with other records at the same name, and it can't sit at the zone apex. That's why Route 53 has alias records.

2. TCP: reliable delivery

  • Three-way handshake: SYN → SYN/ACK → ACK. Only then does data flow.
  • A port identifies the service: 22 SSH, 80 HTTP, 443 HTTPS, 53 DNS, 3306 MySQL, 5432 PostgreSQL.
  • TCP vs UDP: TCP = ordered, acknowledged, retransmitted. UDP = fire and forget (DNS queries, streaming, QUIC builds on it).
  • Symptoms to recognize:
    • Connection refused → host reachable, nothing listening on that port.
    • Timeout → packets dropped silently (firewall / security group / wrong route).

That distinction (refused vs timeout) will save you hours in AWS later.

3. TLS: the encrypted channel

  1. Client sends ClientHello with SNI (the hostname it wants, so one IP can host many sites).
  2. Server returns its certificate chain (leaf → intermediate → root CA).
  3. Client verifies: chain trusted? hostname matches? not expired?
  4. Both sides derive session keys; everything after is encrypted.

What a certificate proves: whoever answered controls that domain name, as validated by a CA.
What it does NOT prove: that the site is honest or safe. Phishing sites have valid certificates too.

4. HTTP: the conversation

Code Meaning Think
200 OK worked
301 / 302 redirect follow Location header
400 bad request client sent garbage
401 unauthenticated who are you?
403 forbidden I know you, no access
404 not found wrong path
429 too many requests rate limited
500 server error app crashed
502 / 503 / 504 bad gateway / unavailable / gateway timeout typical load balancer symptoms

Methods: GET read, POST create, PUT replace, PATCH modify, DELETE remove.
Headers worth knowing: Host, Cookie, Set-Cookie, Authorization, Content-Type, Location, Strict-Transport-Security.

5. IPv4 and CIDR: do it by hand

/N means the first N bits are the network part. Addresses in the block = 2^(32−N).

CIDR Addresses
/16 65,536
/24 256
/26 64
/28 16

Private ranges (RFC 1918): 10.0.0.0/8, 172.16.0.0/12, 192.168.0.0/16.

Worked example A: 192.168.10.77/26

  • Block size = 64 (since /26 leaves 6 host bits, 2^6 = 64).
  • 77 falls in the block starting at 64 → network = 192.168.10.64
  • Broadcast = 192.168.10.127 (64 + 64 − 1)
  • Usable hosts = .65 – .126 (62 hosts)

Worked example B: 10.0.5.200/22

  • /22 → block size 4 in the third octet.
  • 5 falls in the block starting at 4 → network = 10.0.4.0
  • Broadcast = 10.0.7.255
  • Usable hosts = 10.0.4.1 – 10.0.7.254 (1,022 hosts)

AWS preview: AWS reserves 5 addresses in every subnet, so a /24 subnet gives 251 usable IPs, not 254. We'll use this in the VPC chapter.

Quick drill (answers at the bottom)

  1. How many addresses in a /20?
  2. What's the broadcast of 172.16.33.9/21?

6. Routing, NAT, stateful vs stateless

  • Default gateway: where your host sends anything not on its local network.
  • NAT: many private hosts share one public IP; the NAT device rewrites addresses and tracks connections.
  • Stateful filter: remembers connections; if outbound is allowed, the reply is automatically allowed. (AWS security groups.)
  • Stateless filter: judges every packet alone; you must allow return traffic explicitly. (AWS network ACLs.)

Build: trace one request

Run these against any site and label each output on your diagram:

dig +trace example.com # full DNS resolution chain curl -v https://example.com # TCP + TLS + HTTP in one view openssl s_client -connect example.com:443 -servername example.com # cert chain ss -tulpn # what's listening on YOUR machine traceroute example.com # path through routers 

Deliverable: redraw the request diagram above from memory (draw.io / Excalidraw / paper) with the real output of each command next to its layer. Drawing it yourself is what makes it stick.

Break it (and diagnose from the CLI)

Break How (in a lab / local machine) How to diagnose
Wrong DNS add 127.0.0.1 mysite.test vs a wrong IP in /etc/hosts dig, getent hosts, curl -v shows the IP used
Blocked port 443 local firewall rule dropping 443 curl -v hangs (timeout), nc -zv host 443
Nothing listening stop your local web server curl → connection refused, ss -tulpn shows no listener
Expired / mismatched cert use curl https://expired.badssl.com or wrong.host.badssl.com (public test sites) openssl s_client + curl -v error text

Write each fix as a 3-line runbook: symptom → check → fix.

Cheat sheet

Symptom Likely layer First command
"Could not resolve host" DNS dig example.com
Hangs then times out Firewall / routing nc -zv host port, traceroute
Connection refused Nothing listening ss -tulpn on the server
Certificate error TLS openssl s_client -connect host:443 -servername host
403 / 404 / 5xx HTTP / app curl -v, check app + LB logs

✔ Checkpoint

You're ready for the next chapter when you can, without notes:

  • explain DNS → TCP → TLS → HTTP in order,
  • compute network/broadcast/host range for any CIDR,
  • tell "refused" from "timeout" and know what each implies,
  • diagnose a bad cert from the command line.

Drill answers: 1) 4,096. 2) /21 → block size 8 in the third octet; 33 is in 32–39 → broadcast 172.16.39.255.

📚 Docs

Next up: Chapter 2, Linux for cloud engineers (users, permissions, systemd, logs, SSH hardening).

Full roadmap with every phase, certification path and job mapping:
👉 Read the complete AWS Cloud Security Roadmap on Medium

Lab note: commands tested conceptually against public test domains only. Only test systems you own or are authorized to test.