Trezor/Keepkey = 100% open source, but more vulnerable to physical attacks on its generic controller to extract plaintext mnemonic. (mitigated by use of passphrase)
Ledger Nano/HW.1 = partially closed-source smartcard element to store plaintext seed (no passphrase option), on which physical attacks are much more difficult than on a generic controller, but possibly backdoored?
that's about right - just consider that the STM32 could also be backdoored at a lower level. That's a common issue with hardware, you have no way to be sure unless you build it yourself, which is not possible to achieve.
Also, btchip can you explain more details about how the mobile phone second-factor works? Is the pairing with the Security Card performed only once, or for each transaction?
The pairing is performed once when the firmware is initialized.
I assume it is not actual multi-sig, but rather an internal security function built into the chrome app? If the Ledger chrome app gets compromised, couldn't it then display the same malicious transaction details on your computer and on the phone app?
The algorithm is described
here - the chrome app just forwards the encrypted transaction details to the phone, which decrypts it.
So a malware would need to compromise both and keep synchronized to display the same information on the desktop computer and the phone.
