What Actually Happened

The attack wasn't a technical exploit in the traditional sense. Nation-state adversaries didn't find a zero-day vulnerability in the blockchain protocol itself. Instead, they went after the humans holding the keys.

They deployed a malicious TestFlight app and exploited vulnerabilities in development tools like VSCode and Cursor. They social-engineered signers into pre-signing transactions they didn't fully understand, using durable nonce transactions to hide what was actually being authorized. And they worked the system itself, getting Drift's Security Council migrated to a 2/5 threshold with zero timelock, eliminating any detection window before execution.

The result: signers thought they were doing one thing while actually authorizing something completely different.

The Fundamental Problem with Multisig

Here's what multisig actually does: it distributes trust across N humans, but it doesn't eliminate the human attack surface. It multiplies it.

Every signer is a potential entry point. Every device holding key material is a target. Every social interaction is an opportunity for manipulation. When you're up against sophisticated adversaries with time, money, and resources, both the device security and social engineering assumptions that multisig relies upon can be defeated.

The keys still exist. They sit on signing devices, waiting to be used. And as we've seen across $3.1B in thefts during H1 2025 alone, when keys exist somewhere, motivated attackers find ways to compromise them.

A Different Architecture Entirely

Lokblok works on a completely different principle. Phantom Secrets technology, covered under U.S. Patent No. 12,438,716 B2, means keys don't exist on signing devices at all.

Instead, key material is reconstructed ephemerally inside HSM boundaries only at the exact moment it's needed for a cryptographic operation, then immediately destroyed. There's no standing secret to extract from a signer's device, no durable key material to social-engineer someone into misusing.

In the Drift attack scenario, even if adversaries had successfully compromised every signing device and convinced every signer to authorize transactions they didn't fully understand, there would have been nothing to extract. The malicious TestFlight app would find no key material. The VSCode exploit would access no secrets. The social engineering would fail to produce persistent keys that could be manipulated.

You can't steal what isn't there over time.

Why This Matters Now

The Drift breach, like the $1.4B Bybit theft and countless others, shows that following apparent best practices isn't enough anymore. Multisig was supposed to be the answer. Hardware security modules were supposed to solve the problem. Air-gapped devices were supposed to keep secrets safe.

But 100% of major breaches involve key compromise, because all of these approaches still assume that if you have to keep something safe, it has to exist somewhere.

We've built Lokblok around the understanding that this assumption is what everyone else gets wrong. Secrets don't need to persist. They shouldn't. When there's no standing secret to target, the entire attack surface that enabled Drift's breach simply disappears.

That's not an incremental improvement. That's a completely different underlying architecture and ethos for how cryptographic operations should work when the adversary has nation-state capabilities.