09/13 Notes Chapter 1

Recognition

Either:

• verification
• identification

Verification

• 1 to 1 matching

• identity is clamed

  • i.e. Iphone verifying owner fingerprint (1 identity)

• output is yes or no (genunine or imposter)

Identification

• 1 to N matching

• Asks system “Who am I?”

  • No identity is clamed

• Output is Identity or no match (not binary since multiple ids)

De-Duplication

Avoid duplication of identities

• similar to identification

Traits

Trait should ideally be

1. Unique or distinctive
2. Easy to collect
3. Permanence (does it completely change?)
4. Universality (most humans have this trait)

5. Integrity (is it easy to replicate?)

   • Trait is resiliant to circumvention

     • Spoofing and obfuscating

6. Performance (how easy is it to get information from the trait?)
7. Acceptability (how willing is population to give trait?)

System / Applications

1. Cooperative vs Non-cooperative user(bank user vs criminal at station)
2. Overt vs Covert Deployment
3. Habituated vs Non-Habituated users (used to getting data collected or not)
4. Attended vs Unattended operation (does the user require human interaction)
5. Control vs Uncontrolled (environment is very controlled vs not)
6. Closed vs Open (data is not shared with another application or is shared)