Trading money for stuff at a distance
Hard part: handling the money
Players in a transaction
[customer] <----------> [merchant]
\ /
\ /
[ bank(s) ]
Online (3-party, includes bank) vs offline (2-party, no bank). Credit card (online) and cash transfer (offline).
Can't prevent entirely, can only manage it (actually, there's an optimal level of fraud, where the marginal cost = marginal benefit. Sometimes gets too expensive to prevent more fraud: spend $1million to prevent $1000 in fraud? ).
Want fraud to be:
- low % of total transcation value
- spread "fairly" (all users pay a little, i.e. pay first $50 of fradulent charges. If $10M in fraud, spread among 100M users (not 10!). Don't scare people from getting card). Sometimes more economical to allow some fraud.
- paid for by party who can prevent it (good principle in general)
- incentive to prevent fraud: merchange eats cost
- ATM cards: in US, if disputed transaction, bank ate cost. In UK, if disputed, customer ate cost. Result: in US, banks put in security cameras, and fraud dropped. In UK, customers couldn't prevent fraud.
- Economics: make person who can make change feel the burn
- Bank issues card number to individual (card easy to duplicate)
- Customer, merchant meet in person
- Merchant calls bank to check card status (credit limit - velocity control. Check if card valid, not stolen)
- Bank responsible for collection
- merchant eats some fraud costs
loss management techniques
- credit limit
- AI-based scanning of transaction stream; call customer if anything odd noted.
- customer incentive to challenge bad charges
Characteristics of fraud
Chip on card; stable memory; powered externally
Stronger authentication of the card. Not really worht cost at present (cost a few dollars/card... and that's how much fraud costs. Would have to drive fraud to near zero to be economical).
Different elsewhere in the world
- US has cheap, reliable phone service
- smartcards have some offline checks
Key space not large enough
- each card used once -- can't double charge
- but what it provides (anonymity), customer's don't really want
Main diff from in-person: no signature from customer.
Safeguard: ship to billing address of end
Merchants take calculated risks
When selling digital bits, fraud likely. Sell software by download: less common now.
On the net: use SSL, mimic phone transaction
- prev SECSEP protocol. Advanced, but slow. Not adapted
- client-side certificates never took off. $10-20 per year, tough to use, customer doesn't care about fraud.
FICTION: signatures are not forgeable
- people rarely check
- can challenge transaction: store shows signature. Reminds honest person.
- digital signature, with those special pads: track pressure, timing, not just image itself.
