Network Security Basics: Encryption, Firewalls, Attacks

Every protocol in the earlier notes was designed for a network of trusted researchers: IP does not check source addresses, ARP believes any reply, and classic DNS and HTTP send everything in clear text. Network security adds what those protocols lack. Cryptography keeps data secret and tamper-evident, certificates prove who is on the other end, firewalls and VPNs control what may cross a boundary, and specific defences blunt specific attacks.

The CIA triad

Three more terms are often added: authentication (proving who you are), authorisation (what you are allowed to do) and non-repudiation (a sender cannot later deny sending a message, provided by digital signatures).

Symmetric encryption

One shared secret key both encrypts and decrypts. It is fast, often built into processor hardware, so it encrypts all bulk data. AES (block size 128 bits, keys of 128, 192 or 256 bits) is the standard; ChaCha20 is a fast alternative in software. DES, with its 56-bit key, can be brute-forced and is broken, and 3DES is deprecated.

Its weakness is key distribution: the two parties must already share the key secretly, and every pair of people needs its own key, n(n − 1)/2 of them for n people.

Asymmetric (public-key) encryption

Each party has a key pair. The public key can be published; the private key never leaves its owner. What one key does, only the other can undo:

• encrypt with the recipient's public key, and only the recipient's private key can decrypt (confidentiality);
• sign with your private key, and anyone can verify with your public key (authenticity).

1. Symmetric: Alice and Bob share one secret key. A toy XOR cipher turns 0100 into 1111 and the same key turns it back. Fast, but the key had to reach Bob secretly first.
2. Asymmetric: Alice encrypts with Bob's public key, 4^3 mod 33 = 31. Only Bob's private key undoes it: 31^7 mod 33 = 4. The public key could have been printed in a newspaper.

RSA, whose security rests on the difficulty of factoring a large number, and elliptic-curve algorithms (ECDH, ECDSA) are the common families. They are far slower than symmetric ciphers, so they are used for small jobs: agreeing keys and signing.

Worked example: toy RSA

Real RSA uses primes hundreds of digits long; small numbers show the mechanics.

Anyone may know (3, 33), but finding d = 7 requires φ(33), which requires factoring 33. For a 2,048-bit n, no known method can factor it in practice.

Worked example: Diffie-Hellman key agreement

Diffie-Hellman lets two parties agree a secret over an open channel without ever sending it. Public values: prime p = 23, generator g = 5.

1. Alice and Bob agree in public on a prime p = 23 and a generator g = 5. Anyone listening, Eve included, knows both.
2. Each picks a private number and never sends it: Alice 6, Bob 15. Each raises g to it: Alice gets 5^6 mod 23 = 8, Bob 5^15 mod 23 = 19.
3. They swap A = 8 and B = 19 in the open. Eve now holds 23, 5, 8 and 19, but turning 8 back into 6 means solving a discrete logarithm, which is infeasible for real key sizes.
4. Alice computes 19^6 mod 23 = 2 and Bob 8^15 mod 23 = 2: the same number, since both equal 5^(6 × 15) mod 23. The shared secret 2 never crossed the network.

TLS 1.3 uses the elliptic-curve version with fresh numbers for every connection (ephemeral keys), which gives forward secrecy.

Symmetric vs asymmetric

Real systems are hybrid: TLS uses asymmetric cryptography to authenticate the server and agree a key, then encrypts the data with a symmetric cipher such as AES-GCM (HTTP and HTTPS).

Hashing

A cryptographic hash function turns any input into a fixed-size digest. SHA-256 always outputs 256 bits (64 hexadecimal characters). A good hash is deterministic and fast, and it is:

• preimage resistant: given a digest, you cannot find an input that produces it;
• second-preimage resistant: given an input, you cannot find another with the same digest;
• collision resistant: you cannot find any two inputs with the same digest;
• sensitive to change: flipping one input bit changes about half the output bits (the avalanche effect).

MD5 and SHA-1 have practical collision attacks and must not be used for security; SHA-256 and SHA-3 are current. Passwords are stored as the output of a deliberately slow, salted function such as bcrypt, scrypt or Argon2: the random salt per user means identical passwords get different hashes and precomputed tables are useless. An HMAC mixes a secret key into a hash, so the receiver can check both integrity and that the sender knew the key.

Digital signatures and certificates

A digital signature is a value computed over the message's hash with the sender's private key, which anyone can check with the sender's public key:

1. The sender hashes "PAY 50" to 4 and raises it to the private exponent: 4^7 mod 33 = 16. The signature travels with the message, which stays perfectly readable.
2. The receiver hashes the message itself and raises the signature to the public exponent: 16^3 mod 33 = 4. The two match, so the message is unchanged and only the private key's holder could have signed it.
3. Change the message to "PAY 90" on the way and its hash becomes 8, while the signature still opens to 4. They differ, so the forgery is caught; making a new signature would need the private key.

A valid signature proves integrity (any change breaks it), authenticity (only the private key holder could make it) and non-repudiation. It does not hide the message.

But how do you know a public key really belongs to example.com? A digital certificate (X.509) binds a public key to a name, and a certificate authority (CA) signs it. Browsers and operating systems ship a trust store of root CA certificates; a site's certificate is signed by an intermediate CA, which is signed by a root, forming a chain of trust. A client checks each signature up the chain, the validity dates, that the certificate's names include the host, and that it is not revoked (through CRLs or OCSP).

Firewalls

A firewall enforces rules about what traffic may cross a boundary.

Next-generation firewalls combine stateful filtering with application identification and intrusion prevention, and a web application firewall (WAF) sits in front of a website to block attacks such as SQL injection and cross-site scripting. Rules are checked top to bottom and the first match wins:

1. HTTPS to the web server from anywhere: rule 1 matches at once and the packet is allowed. The rules below it are never looked at.
2. SSH from 10.4.2.9, an internal address: rule 1 (wrong port) does not match, but rule 2 does, because 10.4.2.9 is inside 10.0.0.0/8. Allowed.
3. SSH from outside, 198.51.100.7: rule 1 (wrong port) and rule 2 (wrong source) do not match, so it falls through to rule 3, the default deny. Dropped.

VPNs

A virtual private network carries traffic through an encrypted tunnel across an untrusted network, so a remote laptop (remote-access VPN) or a whole branch office (site-to-site VPN) behaves as if it were on the private network. Common protocols are IPsec at the network layer, TLS-based VPNs, and WireGuard. IPsec's ESP protocol provides encryption and integrity (its AH protocol provides integrity only), in transport mode, which protects the payload of the original packet, or tunnel mode, which wraps the entire original packet inside a new one, as gateways do. A VPN hides traffic from the local network and the ISP, but the VPN operator can see it, and it does not stop malware or phishing.

Common attacks and defences

The attack that certificates exist to defeat is the man in the middle:

1. After ARP spoofing on the local network, the client's packets for example.com reach the attacker first. The attacker accepts the connection, hoping to sit in the middle and read everything.
2. To relay, the attacker opens its own connection to the real server and gets the real certificate. It cannot reuse that certificate towards the client: it lacks the matching private key to sign the handshake.
3. So it presents a certificate for example.com that it made itself, signed by its own CA. The client finds no chain from that issuer to any root in its trust store, so the check fails and it aborts before sending any data.

The SYN flood and SYN cookies are explained in TCP and UDP, and DNS cache poisoning in DNS.

Common mistakes

• Saying hashing is encryption: a hash cannot be decrypted; it is one-way.
• Saying a digital signature encrypts the message: it proves who sent it and that it is unchanged, but the message stays readable.
• Mixing up which key does what: encrypt with the recipient's public key; sign with your own private key.
• Treating Base64 as security: it is encoding, reversible by anyone.
• Storing passwords with plain SHA-256: use a salted, deliberately slow function such as bcrypt or Argon2.
• Believing HTTPS or a VPN stops phishing: they protect the channel, not a user who types a password into the wrong site.

Interview questions

Why does TLS use both asymmetric and symmetric encryption? Asymmetric cryptography solves key distribution and proves identity, but it is far too slow for bulk data. So TLS uses it only to authenticate the server and agree a shared key, then encrypts the actual traffic with a fast symmetric cipher such as AES.

How many keys do 100 users need with symmetric and with asymmetric encryption? Symmetric: one per pair, 100 × 99 / 2 = 4,950 keys. Asymmetric: one key pair per user, 200 keys, since anyone can encrypt to a user with that user's single public key.

Why should passwords be hashed with a salt rather than encrypted? A server never needs to recover a password, only to check one, so a one-way hash is enough, and there is no decryption key to steal. A unique random salt makes identical passwords hash differently and defeats precomputed tables, and a slow function makes guessing expensive.

How does a digital signature provide non-repudiation? Only the holder of the private key could have produced a signature that the matching public key verifies. If the public key is bound to a person or company by a certificate, the signer cannot credibly deny having signed, unless the private key was stolen.

What is the difference between a stateless and a stateful firewall? A stateless firewall judges every packet alone by its header fields, so replies need their own broad rules. A stateful firewall tracks connections, automatically allows the replies to connections started from inside, and drops packets that belong to no known connection.

How does HTTPS stop a man-in-the-middle attack? The attacker can intercept the connection but cannot present a valid certificate for the real domain signed by a trusted CA, nor sign the handshake with the real server's private key. The browser's certificate check fails and it refuses to continue. Users clicking through certificate warnings is how such attacks still succeed.

What is a DDoS amplification attack? The attacker sends small requests to open servers, such as DNS resolvers, with the victim's address forged as the source. The servers send much larger replies to the victim, multiplying the attacker's bandwidth. Defences include ISPs filtering spoofed source addresses, closing open resolvers and absorbing traffic with large scrubbing networks.

What is the difference between IPsec transport mode and tunnel mode? Transport mode protects only the payload and keeps the original IP header, used between two hosts. Tunnel mode encrypts the whole original packet, header included, and puts it inside a new packet between two gateways, which is how site-to-site VPNs work.

That completes the Computer Networks notes. Test yourself with the Computer Networks (Basic) skill test, then the Computer Networks (Intermediate) skill test.