notes.md

Designing Secure Systems

Computer Security (27/09/2017)

• Common criteria [ISO15408]
  • An international standard for computer security certification
• Protecting digital assets from threats

Secure system definition

• Inability for attackers to achieve:
  • Adversarial goal (breaching any of:)
    • Confidentiality
    • Integrity
    • Authenticity
    • Availability
    • Accountability
  • Means of resources of the adversary
  • Access to the system

Attacker motivation

• Profits and other benefits
• Political activism, terrorism
• Enjoyment
• Development of science and offensive technology

Why things happen

• Bugs or not caring
• Programming is done with an absence of security considerations
• Economics/business
  • customers do not see/care about security
  • security measures sometimes reduce usability

Risk management

• Measuring or/and assessing risks
• Developing strategies and solutions to manage risks:
  • reduce/avoid and handle risks

Residual risk

What remains after defences are in place.

Defenders

• Prevent
  • Deter
  • Hinder
• Detect
  • Monitoring/logging
  • Anomaly analysis
• Respond
  • Incident management
  • Forensics
  • Change procedures
  • Install new technologies

Attack trees (04/10/2017)

• Formal analysis of all known attack venues
• Types of nodes
  • OR nodes (look like an arrow)
  • AND nodes (like an arrow but with a semi-circle inside)
• Labelled with probabilities or cost estimates
• Main goal is at the top

Secrecy

• Create entry barriers for competitors
• Defend against hackers

Kerckhoff's principle

The system must retain secure should it fall in enemy hands..

Responsible disclosure

• Researchers should disclose vulnerabilities to the system owners and give "reasonable time" to fix them
• Benefits
  • Creates incentives for fixing vulnerabilities
  • Companies advertise bounties

Principles of secure design (11/10/2017)

12 principles

• Inspired by simplicity and restriction
• Simplicity:
  • Makes designs/mechanisms easy to understand
  • Less to go wrong with simple designs
• Restriction:
  • Minimises power of an entity
  • Can only access info it needs
  • Can only communicate with other entities when necessary

Principles

1. Secure the weakest link
   • Security only as strong as weakest link
   • Attackers go for weakest point in a system
2. Defence in depth
   • Defend a system using several independent methods
   • Redundancy and layering to not rely on one defence
   • Consider: people, technology, operations
3. Fail secure
   • For a type of failure, access or data are denied
4. Grant least privilege
   • Every module can only access what it needs
   • Separation of privileges
     • Split system into pieces, each with limited privileges
   • Segregation of duties
     • Hard for one person to compromise security
5. Economise mechanism
   • Avoid overly complicating a system
   • Complexity leads to insecurity
     • Designer can't keep up
     • Harder to analyse security
6. Authenticate requests
   • Assume that operating environment is hostile
   • Put checks to ensure dependencies haven't been compromised/spoofed
   • Anticipate command-injection, XSS, etc.
7. Control access
   • Accesses should be checked
   • Up-to-date permission checking (not caching authority granting results)
8. Assume secrets not safe
   • An attacker will know how the system works
   • An attacker can find keys in binaries, use tools at disposal, etc.
9. Make security usable
   • Annoying security/product will be avoided by users
10. Promote privacy
    • Only collect personal info required
    • Store personal info securely, limit access
    • Delete personal data once purpose is served
    • Only store encrypted data
11. Audit and monitor
    • Record what actions were performed and by who
      • Disaster recovery and accountability
12. Proportionality principle
    • Maximise utility vs maximise security?
    • Maximise utility while limiting risks
      • to acceptable level
      • within reasonable cost

Trusted: Something that could break security policy Trustworthy: Something that will not break security policy

Cryptography (11/10/2017)

Symmetric key encryption (e.g. AES)

• Encryption key used to encrypt and decrypt
• Typically use a block cipher
• Block modes such as "counter-mode"
• Needs randomness for security

Hash functions (e.g. SHA2)

• Takes a message of any size and outputs a hash value of fixed length
• H(x) = H(y) implies x = y with huge probability

Public-key encryption (e.g. RSA-OAEP)

• Public key (encrypt) and secret key (decrypt)
• Very difficult to retrieve secret from public
• Allows building encryption without needing key agreement

Hybrid encryption

• No block modes for public key encryption

  • Would be too slow for long messages

• To encrypt long message m using public key pk:

  • Choose random symmetric key, k
  • Encrypt m with k: c1 = enc(k, r1, m)
  • Encrypt k with pk: c2 = enc(pk, r2, k)
  • Send c1, r1, c2

Digital signatures (e.g. Schnorr)

• Public and secret key
• Used for authentication (signatures)
• Signature checking algorithm only returns true if signature used correct secret key

Authenticating websites and encrypting web traffic

Server has public key pk

1. Client->Server: "hello", random rb
2. Server->Client: pk, random rs
3. Client->Server: random pms, encrypted with pk

Certificate authority model

TLS

• Server sends public key
• Session encryption key established from server public key

Certificate authorities

• Trusted third party that asserts that a public key belongs to a website
  • CA signs certificate to do this
• Browser knows the verification key to verify a public key's signature
• CA verification keys built into browser

Device Security (18/10/2017)

• Physical security

• Firmware/OS security

• Application security

• Protects devices against

  • Malicious applications
  • Rootkits

Malicious applications

Distributed using OS specific applications, designed to exploit the operating system vulnerabilities

• High success rate, masquerading as useful applications
• Often used to install more dangerous malware (e.g. backdoors, rootkits)

{Missing notes}

How to prevent a rollback attack? (25/10/2017)

• Counter based version control
• Blacklist version control
• eFuses (physically blow when firmware is updated)
• Apple nonce based protocol: random unique value generated at every restore, signed by Apple

Data Security

Key encapsulation

• Encryption of large data slow with public key encryption
• Steps:
  • Encrypt data with symmetric key encryption
  • Encrypt symmetric key with public key
  • Send both encrypted data

What you get/don't get from encryption

• Does:
  • Protect data while resting
  • Protect data from apps that don't have access to the keys
  • Protect data if stolen or accessed
• Doesn't:
  • Prevent data loss
  • Make the system more resilient
  • {Missing point}

Disk based encryption

(Stuff about what's encrypted and what's not.. e.g. kernel not encrypted)

Storing the key

• USB stick
  • Easy
  • Requires USB on system
  • Vulnerable to stealing
• TPM
  • Difficult to set up
  • Transparent
  • Protected from stealing

{Missing notes}

Bitlocker

{Missing notes}

Cloud Storage Security

{Missing notes}

Metadata Security

Cryptography issues

• Protecting metadata, using public key crypto gives larger attack surface
  • Symmetric doesn't have this problem and is more efficient
• Symmetric keys are hard to share
  • Use public key exchange protocols to share

We want:

• Content security: data is encrypted
• Metadata security: ownership info, timestamps, access rights, ciphertext length, etc.
• Access pattern security: when accessed, who accesses, how accessed, etc.

Searchable encryption

We can decrypt emails with key, but can the email provider? Solution: Searchable encryption.

• Generate searchable index (precomputed search)
  • Foreach(doc in DB.docs){ foreach(word in doc) { if (word == keyword) { store(doc.id) } } }
• Store index encrypted in the cloud

Forward index

List keywords associated with documents

• Efficiency:
  • As no. of docs increases, so does time
  • As no. of keywords increases, so does time

Inverted index

List documents associated with keywords

• Efficiency:
  • As no. of keywords increases, so does time

{ Missing notes}