Fear the reaper: characterization and fast detection of card skimmers

I don't know what you mean by "ordinary" 2FA, but it's pretty common in Thailand when making online payments to have an SMS sent to you, and in the UK my Amex card has been known to require SMS/email codes for online purchases.

Will be phased out soon due to regulation (PSD2) as it doesn't meet the requirement for 2 factors.

Some banks also allow you to confirm the transaction inside their mobile app, which is pretty convenient.

Inserting the whole card is also a security measure to be able to take the card away if an incorrect pin is entered 3 times, or if the card is left in the device after leaving the ATM.

Countries that have completed EMV roll out still issue cards with mag stripes. It's maybe worth briefly explaining why this might not matter at all:

The payment card networks have two _entirely separate_ mechanisms in play. To a Hacker News reader it may seem insane to not tie them together, but this is how they evolved and apparently banks would rather eat fraud and error costs than spend whatever it would cost to fix.

1. Authorisation. This mechanism is in charge of deciding whether all the parties are OK with a transaction, e.g. "Mike Smith pays Example Co. $145". Cryptographic security can be used to verify that this was really Mike Smith's authentic card, he entered his PIN, Example Co asked for exactly $145 and Mike's bank said that was OK.

2. Payment. This operates entirely on the honour system between banks and retailers. Example Co. says Mike Smith agreed to pay $145 and the backend systems automatically give $145 of Mike Smith's money to Example Co. Done.

Because these two systems aren't tied together at all, even though the authorisation system guarantees the exact amounts, is strongly replay resistant, and so on, none of those properties apply to actually paying your money to some retailer. So regardless of whether it's tricky to get past authorisation it doesn't matter to big fraudsters, they can just take whatever they want and ignore the Authorisation step altogether.

When you realise somebody took your money you will call your bank. They may try to give you the run around (after all, it's easier) but ultimately - perhaps after some notice period - they'll probably agree to reverse that clearly fraudulent charge and if you live in a country with actual consumer protection they may even have to pay back any fees or other financial penalties you incurred as a result of them accepting the fraudulent payment request.

I'm pretty sure the swipe-style is aggresively deprecated by the banks. I talked to a vendor who had an older terminal that would not accept NFC, and he mentioned that within 2 years, he'd have to get a terminal that would process NFC payments.

I don't know the details of the technological differences in transaction communication between the two, but in the UK chip and pin is noticeably and consistently faster to perform transactions in my experience, often to the the point that it's perceptibly instant... although it has been quite some years since i've used magnetic ones so that's from memory.

I suppose one way to protect yourself if you are never going to use your card in the US, is by destroying the magnetic strip (magnetically).

When I'm at O'Reilly Auto Parts, the pinpad is beeping at me to remove my card almost the moment I've inserted it. It's shockingly fast. (Then their neolithic-age printer takes 8 more seconds to generate a receipt, but... baby steps!)

I wish whoever set up their system could go show some others how it's done. Because yes, on average, most retailers have abominably slow processing for chip cards.

Also, I don't think there are any places left where you swipe the magnetic stripe. If for some reason you don't use contactless, it's a chip transaction. I believe it's pretty much the same all across Europe?

It makes a much more recently deployed system in my area of the US look like a total embarrassment: seconds per processed tap, a much higher tap failure rate (regularly see tap failures for myself and others), etc.

An example I know about with one of those flaws was the "Yes card" which MITMs a real (presumably stolen) card and arranges that the conversation goes like this:

* Legitimate terminal "Hi, I'm a Legitimate terminal, who are you?"

* Real card: "Hi, I'm Sizzle's Real Payment Card from Real Bank"

* Legitimate terminal "OK, let's do an offline transaction. I want Sizzle to authorise payment of 24.50. They entered PIN 1234, is that OK?"

MITM Yes card blocks this and tells the card instead:

"OK, let's do an offline transaction. I want Sizzle to authorize payment of 24.50, but I can't be bothered with a PIN so let's skip that"

* Real card: "OK, yes, payment of 24.50 sounds fine" (cryptographically signed message)

So this way you don't need the correct PIN, since the card never realises you entered a PIN, and so the transaction OK simply never mentions the PIN at all. You don't clone the card though, you need a real card, you're just using something like a Confused Deputy attack where the card and terminal misunderstand the situation.

A _smart_ backend at the bank could identify this fraud when the offline transactions are processed, hours or days later, but many did not, and even if they did spot it the fraudsters got away with their transaction meanwhile. The permanent "fix" for this was to roll out new cards and terminals, given this costs money it was probably not done widely or quickly.

If they [the ATMs] do this and can read the card, then they can also check that the measured movement matches what the controller sent to the motor driver. Heck, depending on the driver they could just let it measure back EMF (e.g. some Trinamic stepper drivers can do that).

Now, if they don't need the full mag stripe, then you've got a problem. But I don't think that mechanism would work for most skimmers.

But for the Authorization a PIN absolutely can be required for online card transactions using a mag stripe. And this was routinely done in countries which had PINs for debit cards years before EMV. The terminal calls the bank and says I have this card, and here's the PIN entered by the customer, is that OK?

Unlike the mag stripe you can't clone a card using the "wireless" EMV mode, the chip isn't just playing back a fixed data stream, it's an active component.

[ It might be instructive to expand here, so I shall ]

A part of an EMV transaction the terminal and card are supposed to pick random ("unpredictable") numbers each time. If this is done correctly it presents a significant security feature. For example, suppose tomorrow I plan to tell a jewellery store's payment terminal that your card, which I was able to access briefly today when you were in the same elevator as me, authorises purchase of a $500 watch, then I'll pawn the watch and keep the cash. Well, I need the cryptographically signed message from your card saying this is authorized. But, that message needs the unpredictable number that the jewellery store terminal will choose tomorrow, which I don't know yet, so I can't do it.

Now, in practice researchers found some terminals and cards are crap and e.g. the numbers they use aren't truly unpredictable. But that's an implementation flaw that can be fixed, just the same as if your bank has a habit of leaving the back door open and the vault unlocked. It's something your defence attorney should know if the bank accuses you of fraud for someone else's transaction, but it's not an inherent problem in "wireless" EMV.

Certainly, using rather basic NFC smart card technology, all but on-line attacks could be eliminated. My question is then, what kind of low-protection protocol do they use in practice to make this so insecure.

Specifically, I am asking about an offline attack that allows an actual spend the bank would accept. I am also only interested in debit cards (because that is what I have) so just reading a CC number from NFC doesn't bother me.