Both modes keep your private key inside the signing device. Neither one hands the key to
your laptop. What differs is the number of ways a compromised computer can interact with
that device, and how much of the interaction you can see.
What a USB connection actually is
A cable is a two-way data path. Your computer speaks a protocol to the device, asks it
for public keys, sends it a transaction to sign, and receives the signature back. That
path is exactly what makes the experience pleasant — Sparrow can discover your accounts,
read the extended public key and push a transaction across in one click. It is also a
standing channel between an internet-connected machine and the thing holding your keys.
The threat this defends against is straightforward and it defends against it well: even
a fully compromised computer cannot extract the key, because the device never sends it.
What USB does not remove is everything that happens around the signature. Malware on the
host can swap the destination address before you approve, can misreport an amount, or can
keep asking for signatures on transactions you did not build. That is why the device
screen exists, and why it is not optional.
What airgapping removes
An airgapped device never connects to the computer. Sparrow writes an unsigned
transaction to an SD card, or renders it as QR codes on the screen; the device reads it,
displays what it is about to sign, produces a signature and sends it back the same way.
No driver, no protocol negotiation, no live session. The only thing crossing the gap is
transaction data you could print out and read.
That closes an entire class of problem: firmware-level attacks delivered over the wire,
vendor software you did not intend to install, a hostile USB controller, or a device that
silently accepts a command you never issued. It also makes the boundary auditable. You
can look at the SD card and know exactly what left the machine.
When the extra friction is worth it
Airgapped signing is slower. Every spend becomes a small ritual: export, walk the card or
the QR stream over, verify on the device, walk the signature back, finalise, broadcast.
For a wallet you spend from weekly, that friction is real and it makes people postpone
things they should not postpone, like moving coins off an exchange.
So split the decision by how you use the money. A spending wallet with a balance you
would be annoyed but not ruined to lose: USB is fine, provided you verify addresses on
the device screen. A savings wallet you touch a few times a year, holding an amount that
would change your life: airgap it, because the twenty extra minutes a year cost you
nothing and the attack surface you remove is the one you cannot inspect. If you are
running a 2-of-3 multisig, mixing the two is
also reasonable — one signer connected, the others airgapped, so no single compromise
path reaches a quorum.