6841 Notes

General Info

Lecturer: Richard Buckland
Course email: cs6441@cse.unsw.edu.au
Tutor email: z5367571@ad.unsw.edu.au

Logbooks - 30%
Something Awesome Project - 30%
Exam (take home) - 40%

Security Engineering

The principles and practice of designing and managing complex systems against adversaries in a world of humans, technology, cryptography and regulation.

Cyber security refers to when things go wrong, when things go out of the spec.

Mindset

DAEH

Defender Mindset: A protective approach centered on preventing attacks or mitigating damage. Focuses on safeguarding assets, anticipating threats, and maintaining stability.

Attacker Mindset: A proactive, offensive approach focused on exploiting weaknesses or vulnerabilities. Emphasises initiative and finding ways to overcome defenses.

Engineering Mindset: A problem-solving approach focused on designing practical solutions while following standards (set of rules followed by all engineers), constraints, testing, modelling and optimising.

measurement
skepticism
testing
review
openness/transparency
treatment of errors
standards
professionalism
closing the loop.

Human Mindset: A recognition-based approach that accounts for the presence of people in systems, acknowledging human errors and limitations.

Security Theatre

Security theater is the practice of implementing security measures that are considered to provide the feeling of improved security while doing little or nothing to achieve it.

Physical Security

Physical security focuses on establishing a solid foundation to safeguard systems from threats such as tampering or counterfeiting.

very hard to get physical security right

Phases of an attack

Generally an attack start with recon and planning until a vulnerability is found which can then be exploited.

Recon

Beginning of an attack, goal is find out more information about what is being attacked.

can be passive or active

0 day attacks

0 days refers to a software security flaw that is unknown to the software vendor and has not been patched so it can be exploited by attackers.

Harvard and Von Neumann Architecture

Harvard architecture has two physically seperate pieces of mememory (slower but safer)
Von Neumann architecture was chosen for legacy computers
- same memory stack for both data and control
- two memory items in one physical piece.

Buffer Overflow

https://eli.thegreenplace.net/2011/02/04/where-the-top-of-the-stack-is-on-x86/

A buffer is memory used as temporary location for data in a program.

The stack (LIFO) grows from high addresses to low addresses

Function call process:

Call sum()
Push parameters (values of x and y) to the stack
Push return address of sum to the stack
Push base pointer (end of parent stack) to the stack
Execute sum()

Even if you call a function multiple times (e.g. using 4 lines to draw a square), only one stack frame at a time - only one function is actually called at once

Risk and Trust

Security Patterns

Anti-pattern (complexity): Overly complex systems increase the likelihood of security vulnerabilities
Dual control: Multiple entities need to work together (all keys turned at same time)
Defence in depth: A multi-layered approach to security, implementing multiple security measures to protect systems in case one layer fails.
Weakest Link: The security of a system is only as strong as its weakest component
Single Point of Failure: A component in a system that, if it fails, would cause the entire system to fail, creating a critical vulnerability.
Anti-pattern (Defender Mindset): An overly defensive approach that focuses solely on preventing attacks rather than also considering detection and response strategies.

Analysing Risk (NIST - National Institute of Standards and Technology)

Low likelihood, low impact - we can mostly ignore these risks
Low likelihood, high impact - very difficult to predict and assess
High likelihood, low impact - easier to make predictions about due to history
High likelihood, high impact - very dangerous

Type I and Type II Errors

Type I error (false positive)

This error occurs when a true null hypothesis is incorrectly rejected.

Test indicates patient has a disease when they actually do not

Type II error (false negative)

This error happens when a false null hypothesis is not rejected.

Test fails to detect the presence of the disease in an infected patient.

Cryptography

Cryptography is the process of hiding or coding information so that only the person a message was intended for can read it.

CIA: Confidentially, Integrity, Authentication

Usually require secrets

Security by Obscurity

Improve security by hiding the details/mechanism of the system, but then it makes the system close-sourced and vulnerable to invisible bugs.

Steganography

Steganography is when you have a message but you conceal the existence of the message itself.

weave the message into a rug
blink out the message

Six design principles for military ciphers:

The system must be practically, if not mathematically, indecipherable.
It must not be required to be secret, and it must be able to fall into the hands of the enemy without inconvenience.
Key must be communicable and retainable without the help of written notes, and changeable or modifiable at the will of the correspondents.
It must be applicable to telegraphic correspondence.
Apparatus and documents must be portable, and its usage and function must not require the concourse of several people.
The system must be easy to use, requiring neither mental strain nor the knowledge of a long series of rules to observe.

Security by Obscurity: Improve security by hiding the details/mechanism of the system (assuming people do not know certain things)

assuming people do not know the frequency of the phone.
should never be used as the only method of keeping your system secure

SQL Injection

SQL Injection is when an attacker can modify the data used in an SQL query such that it is executed as code.

Attacks aim to manipulate the input string into commands, to gain unauthorised access.

Confidentiality

Information measure

An information measure refers to a quantity that calculates the length of the shortest message needed to transmit information.

Example: compressing a file by removing redudancy (repeated parts) and unused space

Growth Rate

Brute Forcing Passwords

Note: $10^9 \approx 2^{30}$

Example:

Length of password: 3, only upper/lower characters: 52
So overall combinations: $52^3 \approx 64^3 = (2^6)^3 = 2^{18}$
On average you find the password after trying half the combinations so $2^{18} / 2 = 2^{17}$ combinations required to find password

<br>

CPU clock speed = 3GHz = $3 * 10^9 \approx 2 * 2^{30} = 2^{31}$ operations per second
4 cores, speed = $2^{31} * 2^2 = 2^{33}$ operations per second
If there are $32 = 2^{5}$ steps per password attempt.
$2^{33}/2^5 = 2^{28}$ attempts per second.

Time to find password: $2^{17} / 2^{28} = 0.0004$ seconds

General rule: $2^{128}$ combinations = $128$ bits enough to protect against brute force (encryption key length not password length).

Codes vs Ciphers

Codes substitute characters for the components of the original message.
Ciphers use algorithms to transform a message into an apparently random string of characters.

Enigma Machine

The Enigma machine is a cipher device that scrambles the 26 letters of the alphabet.

One person enters text on the Enigma's keyboard and another person writes down which of the 26 lights above the keyboard illuminated at each key press.
Entering ciphertext transforms it back into readable plaintext.

Digital Encryption Standard (DES)

Break the message into blocks (e.g. 3 bits), encrypt each block, then put the encrypted blocks together
Repeat this encryption technique
Can still be brute-forced (key is 56 bits)
- As hardware advanced, it became easier and easier to crack
Cracked through differential analysis

Advanced Encryption Standard (AES)

128 bits (10 rounds) 192 bits (12 rounds) 256 bit (14 rounds)

Stream Cipher vs Block Cipher

A stream cipher breaks a plaintext message down into single bits, which then are encrypted individually into ciphertext using key bits.
- Enigma
A block cipher breaks down plaintext messages into fixed-size blocks before encrypting them into ciphertext using a key.
- DES, AES

Side Channels

Indirect leakage of information from a related physical source

Timing attacks, from analysing low level behaviour (taking longer on particular branches than others)
Computer fan speed changing to a certain frequency
Sound coming from typing off keyboard
Radiation leakage from TVs
Pentagon pizza orders spike whenever an operation is ongoing.

Passwords and Keys

Password: A password is used for authentication. It is a piece of information that a user knows and uses to prove their identity to access a system or service.

e.g. hello123

Key: A key is used for encryption or decryption, or for securely accessing data or systems.

e.g. 3AD77BB40D7A3660A89ECAF32466EF97

Human Error

Humans are present at some level in every system and almost every breach is due to human error.

Social Engineering

The most important tool of a hacker is social engineering. Putting individuals into something called "hot state" where you are not thinking clearly.

fishing is targeting groups of people
spear fishing is targetting only individuals

Hardware Security

Software security assumes trustworthy hardware.

Example Risks

Hardware Trojans: hidden malicious modifications on hardware (similar to software trojan)
Reverse Engineering: Used for piracy (to check whether a product is pirated from you, or to copy someone else’s)
IC/IP Piracy and “Overbuilding”: Manufactorer steals design and build more of the product than required to sell.
Counterfeit/forgery: Companies copying products, with cheaper parts that seem as though they have same functionality.

Supply Chains

A supply chain attack is a cyberattack where hackers target vulnerabilities in an organisation's external suppliers or vendors to infiltrate its systems.

Risk Revision

Examples

OceanGate submarine implosion: low probability high impact risks were ignored due to previous successful missions.
Trust exchange: giving the government sensitive data so companies doesnt need to hold sensitive data (government can be compromised).

<br>

The hardest part about risks is spotting the risk, it is invisible and very hard to see.
What are the risks? (you need to find them all the risks)
Risks are not always independent.

Follow a well established framework to analyse risk. (attack trees)

How do we measure risks

A risk has a distribution, normally it is a normal distribution however some risks are fat-tail risks which are more skewed.

Fat-tails risks are hard to deal with
Example: closure of bankestown line, you can't be sure when it will be done.

How do risks interact

Some risks are not independent and some risk are independent, think about correlations between risks.
Example: bushfires $\rightarrow$ cell towers down $\rightarrow$ no communication.

Bits

$10\approx 8 = 2^3$

10 digit number: $10^{10} \approx (2^3)^{10} = 2^{30}$, 30 bits
5 digit number: $10^{5} \approx (2^3)^{5} = 2^{15}$, 16 bits
3 digit number: $10^{3} \approx (2^3)^{3} = 2^{9} \approx$, 10 bits

Every set of $000 \approx 2^{10}$

$2000,000 = 2^1 * 2^{10} * 2^{10} = 2^{21}$
$8000,000,000 = 2^3 * 2^{10} * 2^{10} * 2^{10} = 2^{33}$

Confidentiality

Key Problem

Huge amount of keys to talk to a lot of people confidentially.

$\frac{n^2}{2}$ keys required to talk to $n$ people.

Asymmetric Ciphers

Merkle Puzzle: Imagine A and B wish to communicate securely. A creates 1000 puzzles sends all of them to B. Each puzzle is encrypted, and the content involves a message number and a password. Once B solves one puzzle, he sends the message number to A via plain text (anyone can see, does not matter). So, if anyone else manages to read this message number, it is still safe since they do not know which puzzle it was and which password is related to it.

Symmetric Ciphers: Same key is used to encrypt and decrypt

Asymmetric Ciphers: Two different keys used to encrypt and decrypt

Public encrypt key is shared and private decrypt key is secret.
Works using math functions.

Man in the middle attacks

A Man-in-the-Middle (MITM) attack is when an attacker secretly intercepts and alters communication between two parties.

Terminology

Tranquility

Tranquility refers to the assumption that a system remains in a stable, secure state throughout its operation without change. Roles never change

assuming a student can never become a lecturer

Defence in depth

Defense in depth is a security strategy that involves layering multiple protective measures across different areas of a system or network to reduce the likelihood of a successful attack.

Swiss cheese model

Password Recovery

Use a different email than what is used for work, etc. so that compromised main email does not risk the attacker trying to password reset/recover everything attached to that email.

Social engineering

Timing: Attackers will always attack you when it is least convinient to you, least able to defend, least able to notice.

Attacking during peak period of traffic (more invisible)

Challenge for attackers

get someone to click on a link
get someone to open an attachment

Dating scams

Attacking people’s emotions/vulnerability
Shame attachment, people do not report it

Human Weakness

Cognitive Bias

https://i.insider.com/55ddd6f7bd86ef1e008b6051?width=1200 Mental shortcuts or systematic errors in judgment that can affect decision-making related to security practices and risk assessment

Corruption and self-interest

Beneficial to certain individuals, detrimental to others.

Sometimes you can't see the corruption and self-interest because it is beneficial to some people (bartender calling police to pub, rest of city does not have police).

Firewalls

A firewall is a computer network security system that restricts internet traffic in to, out of, or within a private network.

Attackers will always attack the weakest part of the firewall.

Insiders

Whistleblowers

Whistleblowers are individuals who expose illegal, unethical, or harmful activities within an organisation to authorities or the public.

XSS (cross site scripting)

Original name is cross site scripting because you are able to invoke a script from another website. These days it includes scripts that are injected directly into page for client to render.

allows attackers to run code on client's web browser as it were part of the page itself.

XSS Stored (P1)

Attacker stores malicious code into database through web application. Code is then executed on victim's browser (blog post or bomment)

Attacker stores malicious data into a webserver database
User requests data (like normal)
Database sends malicious data
Malicious data runs on user's site

XSS Reflection (P1/P2)

Parameter sent from user is reflected back onto the page.

Attacker sends malicious link to user
User clicks bad link, and it gets run in the browser
Browser sends stolen data back to attacker

A very common attack, mostly done via phishing. Often done to steal sessions via cookies. <img src="./images/refl.png" width="500" height="auto">

XSS Self

Type of reflected XSS that can only affect yourself.
Not super useful but can mean proper framework/ procedure is not being followed. There may be a vulnerability somewhere else on the application.

Authentication

Authentication has to do with humans whereas confidentiality and integrity refer to the data.

Challenge of authentication: connect physical and digital world for authentication (computer making decisiion does not know anything about outside world)

Privacy

Privacy and security are highly related issues.

Government – want your data for honourable and good services
Companies – want to sell you stuff/make money
You – want your privacy

Privacy can lead to being compromised

Personal information is often used to answer security questions.

Format Strings

A format String is the argument of the format Function and which contains text and format parameters.

printf (“The magic number is: %d\n”, 1911);

The printf() family produce output according to a specific format. With printf, we have specifiers (%d, %s, %x, %n).

%n: write the number of characters written so far into an integer.
printf() is a variadic function, the format strings specifies how many arguments there are.
will take from 1 to an indefinite number of arguments and the format-string specifies how many arguments there are.

grep (global regular expression print): command to search for matching patterns in a file using regex.

Exploiting Format Strings

Demo: https://www.youtube.com/watch?v=NrxbgAGHGqY&list=PLDxcXbiEz9vb1xtzhN4SYd1pfYaAltAZi&index=14 41:00

Extra Resources

Extra Binary Resources:

https://ghidra-sre.org/ - (decompiler - turn assembly to pseudo-- C for reading) https://github.com/pwndbg/pwndbg - (gdb extension - raw gdb is p- ain- - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - )

Format String resources

https://web.archive.org/web/20200324225924/https://www.usna.edu/Users/cs/aviv/classes/si485h/s17/units/06/unit.html https://aeb.win.tue.nl/linux/hh/formats-teso.html

Self Teach:

https://pwn.college/ - (Good site for learning!!) https://github.com/guyinatuxedo/nightmare - (Github repo w a good amt of chals)

Integrity

Data integrity is the assurance that digital information is uncorrupted and can only be accesed or modifited by those authorised to do so.

Blockchains

Not inherently integral since attacks can still occur on blockchains

The Luhn Algorithm

Credit card numbers are generated algorithmically.
Every bank has a unique set of starting numbers.
Bank then generates the rest of the 15 numbers, final digit is generated via the Luhn Algorithm.

Checksum

A way of validating something by storing data into the data itself, useful for checking data intergrity.

Powerful but not all knowing (you can detect an error has occured but not what error)

Types of Hashes

A hash is anything that takes in a variable amount of data and converts it into deterministically generated output of data of fixed length.

Cryptographic hashes

Need to be resistant to certain types of attacks
Slow by design
Examples: MD5, SHA1/2/3
Try to minimise which keys get mapped to the same value

Non-crptographic

Need to be fast and not necessarily unique every time
Example: Creating a hash table in python. If a value is already taken in the table, just use the next cell.

Attacks on Hashing - Pre-image attacks

Given a hash find the plaintext that hashes to the hash

Dictionary attack makes it more likely to find a hit
On average you have to are to find specific instance of 128 bit has after 2^127 tries (half way).

Brute Force

Brute force – go through a whole bunch of common passwords and check if the hashes match.

Attacks on Hashing - Second Pre-image attacks

Given plaintext and its hash find another plain text with same hash.

On average 2^127 tries (half way) to randomly compute a plain text that matches a specific instance of a 128 bit hash AND is not the same as the original text is.

Attacks on Hashing - Collision attacks or Birthday Attacks

Given the hash algorithm find two plaintext that have the same hash.

On average 2^64 hashes must be tried to find a collision.

Attacking Summary

Week 10 movie: Dr StrangeLove (1 Exam question on it)

Cryptographic hash function

Suppose hash(m) is some cryptographic hash function

Pre-image resistance given the hash h, it is computationally infeasible to find the input m_1 such that hash(m_1) = h.

Second pre-image resistance: given a message m_1 it is computationally infeasible to find a second distinct message m_2 such that hash(m_1) = hash(m_2)

Collision resistance: computationally infeasible to find hash(m_1) = hash(m_2) where m_1 is distinct to m_2.

Given those properties we can use cryptographic hash functions as building blocks (primitives) for more complex systems.

Birthday Game

How many people required for a 50% chance that 2 people have the same birthday.

Answer: 23 people $\frac{23 * 22}{2} = 253, \frac{253}{365} = 50$%

Given $n$ people, approximately $\sqrt{n}$ tries to find two people with same birthday.

If you have plain-text that is 128-bit and also a slightly ultered version that is also 128-bit hashed. It takes approx 2^64 little changes before you get the same plain-text value.

Common Hashes

The most common hash right now is SHA2. There are still weaknesses, but SHA2 is better than MD5.

Hashing long messages

Electronic Code Book (ECB)

Break message into chunks, encrypt each chunk and concatenate each chunk together.

Cipher Block Chaining (CBC)

Take the message and break into chunks (P[0], P[1], P[2]).

Counter (CTR)

Break down message into chunks, then take a nonce (number you use once) and a counter and encrypt them together. After XOR the current chunk with the encrypted nonce to get the ciphertext.

Message Authentication Code (MAC)

Imagine you have a message and append some salt to the end and then hash the message. The hashed message is called a MAC = Hash (message || salt).

Length extension attacks

MACs are vulnerable to length extension attacks, where the attacker can use Hash(message1) and the length of message1 to calculate Hash(message1 || bad_message) for an attacker-controlled bad_message, without needing to know the content of message1.

They are able to alter the message without knowing the salt.

Modern day we use a H-MAC.

Safety - V Model of the Systems Development Lifecycle

Engineering principles on the left (safer), testing risker on the right. All that is being tested is the machine is following the right processes with the right level of quality.

Fuzzing testing

Fuzz testing or fuzzing is an automated software testing method that injects invalid, malformed, or unexpected inputs into a system to reveal software defects and vulnerabilities.

GEMS (Generic Error Modelling System) and Errors

Errors can occur at each level of performance:

Skill-based (SB): slips and lapses

usually errors of inattention or misplaced attention

Rule-based (RB): mistakes

usually a result of picking an inappropriate rule
caused by misconstrued view of state, over-zealous pattern matching, frequency gambling, deficient rules

Knowledge-based (KB): mistakes

due to incomplete/inaccurate understanding of system, confirmation bias, overconfidence, availability heuristic (kahneman), cognitive strain.

Humans are reluctant to move from a RB to KB level even if rules aren’t working.

ShatGPT

Amazon S3

Amazon Simple Storage Service (Amazon S3) is an object storage service that is used to share static assets (effective web file share essentially).

Serverless compute and Lambda Functions

Serverless computes means you do not need a server, virtual server, laptop. You do not need to maintain, patch, secure it.

In this case the Lambda function is a very stripped down python docker container that is created every time a user makes a request.

How did they trick openai to give the keys though to bypass SQLi?

User sends input to the AWS Cloud via Chat Web App (Vue) which then sends the input to the AWS API Gateway + Lamda, the Lamda function then sends the input to the OpenAI for it to be summarised. The summarized input is then sent back to the Lambda function, Lamda function then asks the database for results and the results is sent back to the AWS Cloud and it sends back to the user.

When user asks OpenAI, user asks for SQLi injection to be in the payload.

Requirements on how to defend application

WAF rules

Web Application Firewalls that you put in place to allow, block and monitor web requests based on them satisfying certain conditions.

Can be a default; good at detecting generic attacks

User stories

Define requirements for the user. What the user needs and how we get there.

Threat Modelling

Threat modelling is whiteboarding, identify what threats there are, what can go wrong and who can attack.

Primarily come up with solutions for how to solve them

SIEM: Security Information and Event Management

Solution that helps organisations detect, analyse and respond to security threats. Usually in the form of a dashboard.

Splunk, Elastic

Risk based Alerting

Risk objects (Users, Endpoints)
Risk scores (Arbitrary values given to events/other detections)
Risk rules (Risk score exceeded for risk object in last 24 hours)

TLS, SSL, HTTPS

In the old days, emails and information sent across the web was open for everybody to see

Not confidential
None or limited encryption
Had to assume no one else was sniffing
One solution: Encrypt emails before sending and decrypt upon receiving

One-time-pad

Explanation: https://www.youtube.com/watch?v=QVV_bUxxiZ8

Given a plain text of length x, convert it to binary (A). Use a RNG to generate text of length x, convert to binary (B, this is the pad). XOR A and B together (A ^ B = - C) to get encrypted text (C, still length x). The receiver gets C and can decrypt since C ^ B = A- - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - .

Man in the Middle Attack (RSA public private key)

Expection: User A wants to communicate with user B, so they search up B's public key, encrypt their message with it, then send it to user B. B can then decrypt the message with their private key.

MITM Attack: When A searches for B's public key, user C puts their public key there instead. So A encrypts the message with C's public key and sends it off. User C intercepts the message and decrypts it, with their private key (now have access to message). User C then encrypts the message with B's public key and sends it off to B.

In both cases user A and C get the exact same message (no way to tell it was intercepted)

RSA

$M^e$ mod $n = C$

$C^d$ mod $n = M$

$M$ = Message
$e$ = Encryption key
$n$ = constant val
$C$ = Cipher
$d$ = Decryption key

Public key infrastructure (PKI)

Systematic way of getting keys to people, solves the MITM attack.

Company A has a public and private key.

A sends the key and relevant docs to a certificate authority (CA, a trusted register that certifies public keys are valid and actually from company A).
The CA signs the key (hashes it) so and puts it on the internet (called a certificate).
Anyone online can get the certificate and see that it was signed by a CA.

Browsers contain the preinstalled repository of public key of CA's, so to check if certificate signature is valid, they check this list.

Reverse Engineering

https://waugh.zip/6441/re/#/

Finding out how something works by working backwards.

we cover reverse engineering x86 binaries (& how to determine their original source code)

Compilation

compilers are really smart, and will optimize your bad code into good code, so the output doesn’t reflect the input 1:1
you’ll lose context: variable names (not always symbols), comments, #DEFINEs

Tooling

Very important to have tools that perform analysis.

static: read the program without running it
dynamic: show the state of the program as it runs

Static Analysis

disassembling

examine an executable without running it
what instruction do the bytes correspond to?

e.g. ida, ghidra, radare, binary ninja

Dyanmic Analysis

debugging

examine an executable by running it
what are the register values at X point in time
allows you to examine program state as it runs

e.g. gdb, ltrace, strace, windbg

Registers

Stack Frames

x86 uses the cdecl calling convention

parameters are passed on the stack
return value stored in eax register
caller allocates args/eip, callee ebp/locals
callee also cleans up its args

Cleanup Code

function(x, y, z):
push z
push y
push x
call function
add esp, 0xC

Stack

The stack grows from high address to low addresses
- Top of the stack, is lower down in memory
ebp points to the base of the frame (higher in memory)
esp points to the top of the frame (lower in memory)

Variables on the stack

Referenced in relation to ebp (base pointer) e.g. ebp-0x4

local vars are lower in memory (top of stack), so ebp-0x4
arguments are below higher in memory (bottom of stack ) so ebp+0x8

Tamper Proof vs Tamper Evident

Tamper proof: made so that it cannot be interfered with or changed.

safe

Tamper evident: designed to reveal any interference with the contents.

water bottle lid

Dr Strange Love

Characters

General Jack D Ripper – General who has gone crazy and calls the attack
Group Captain Lionel Mandrake – Captain who is trapped with Ripper
Major T. J. "King" Kong – the Major on board the plane which drops the nuclear bomb
General Turgidson - Chairman of the Joint Chiefs of Staff
President Merkin Muffley – president of the USA
Alexei de Sadeski - the Soviet ambassador
Premier Dmitri Kissov – leader of the USSR
Dr Strangelove – Doctor who informs the president of what the Doomsday device is
Colonel “Bat” Guano - Colonel who finds Mandrake

Key Times

3:22 – Mandrake receives the call from Ripper

Do you recognise my voice? Issue: no other authentication done. Could be a recording or impersonation

7:00 - Plane flyer receives code from base and looks into "top secret" book

Lack of encryption, once book is accessed copies can be made easily.
Physical book makes them vulnerable to theft, loss, or unauthorised access.

11:17 - General Turgidson secretary and lover Ms Scott relays a message to him

No authentication from Turgidson, regarding who the message was from and if the messaged she relayed to him was real.

19:25 – Mandrake goes into Ripper’s office with radio

Radio is turned on and has upbeat, "jazz" tune, suggesting that regular radio broadcasts are still live and there is no attack underway.

20:42 – Mandrake realised Ripper has gone mad and tries to call off the attack

He’s locked in the room and cannot communicated with the outside

25:00 – Discussion about why plan R exists

if the chain of command was disrupted a general could issue a nuclear attack
radio channels will switch to a channel where only messages with starting with a 3-letter code can send valid messages.

37:07 - US soldiers fire at other US soldiers

47:00 - First mention of the Doomsday Device

Bomb that will shroud the Earth for 93 years with radiation

48:46 - Dr Strange Love explains the Doomsday Device

Deterance is art of producing the fear of enemies to attack.

57:00 - Mandrake attempts to get the code from Ripper

Ripper realises he has lost and kills himself

1:01 - Missile detotates on plane carrying nuclear bomb

1:02 - Peace on Earth

Mandrake figures out the recall code: encryption

1:03 - Mandrake is held at gunpoint by Guano

No authentication to verify is Mandrake is really who he is
Could just be a Russian in US uniform

1:08 Mandrake attempts to call the president

Guano holds Mandrake at gun point believing he killed Ripper

1:10 Planes with nuclear bombs recalled

OPE code that Mandrake determined and gave to president over phone call

1:15:17 - recalculated fuel

Target changed

1:18:30 - Bomb Safety being released

Multiple users must take safety off (Dual control)

1:23:09 - Nuclear bomb Dropped

Dr Strange Love proposes that everyone goes underground to avoid the radiation

Key Ideas

Single point of failure: Plan R allows General Ripper to gain full control of nuclear weapons
Defence in depth, Insider Attack: Despite all the efforts in authentication General Ripper has bypassed and gained control of the nuclear weapon
Redudancy
Recall code unencrypted
Command and Control strucuture (president, military
Project R should have only been run by the president but, Ripper managed to call it without president knowing
Man in the middle
- Russian ambassador speaking in Russian, trusted position but could have mistranslate
- Russian embassador enters the war room and photographs it, relaying information to Russians
Mutually assured destruction
- If they attack us the response will be so terrible to deter people from attacking, the doomsday machine (Deterance is art of producing the fear of enemies to attack.)
- The missile gap (US got scared because of Russian misinformation)
Sunk cost fallacy: already sent one bomber plane why not go all in
Chain of failure: missile broke the code machine, which resulted in pilots not receiving callback code
military authenticated by uniform
side channels to discover the fallback code (Peace on earth)
Ripper's secret leaked even after death
Type 1 type 2 errors
- false positive: launch second strike w/o first strike
- false neg: not launch second strike when first strike landed
Social engineering between US and Russian president to prevent overreaction
hierarchy of power steep (Ripper overrides president chain of command)
human to human interaction - calls up the russian president even through there is a bad situation
dr strangelove: got super psyched about computing remaining humans with the computer, but exempted themselves

Core Concepts Overview

0-day
Access
Analysis
Availability
Ciphers
Cognitive biases: https://i.insider.com/55ddd6f7bd86ef1e008b6051?width=1200
Command and control (movie)
Complexity
Confidentiality
Culture
DAEH Framework
Defence in Depth
Dual Control
Encryption
Entropy
Errors (TI and TII)
Ethics
Hashing
Humans (in security)
Hubris
Implementations
Insiders
Integrity
Malware
Mindset
Mitigation
OSINT
Privacy
Probability
Ransomware
Risk
Side Channels
Single point of failure
Security by Obscurity
Steganography
Supply Chain
Tamper Evident
Tamper Proof
Threats and assets
Vulnerability
Weakest link
Whistle Blowers