On August 7, 2024, three weeks after South Korea's Virtual Asset User Protection Act took effect, a quiet tender result reshaped the relationship between law enforcement and digital assets. The National Police Agency selected Dunamu—the parent company of Upbit, the nation's largest exchange—to manage seized cryptocurrencies through its subsidiary, Upbit Custody. The contract runs for one year. The news barely rippled through the market. But for those who audit the intersection of code and governance, this is not a footnote. It is a mirror.
Context: The Law, the Platform, and the Custody Gap
South Korea's regulatory framework has matured rapidly. The Virtual Asset User Protection Act, effective July 19, 2024, mandates clearer rules for exchanges, custodians, and user asset protection. Yet enforcement agencies have long struggled with how to securely hold seized digital assets. In the past, police often relied on makeshift cold storage—sometimes even paper wallets—or simply held private keys internally. This introduced risks of loss, theft, or internal malfeasance. The tender for a professional custody service signals a shift from ad-hoc management to institutionalized procedure.
Dunamu, the company behind Upbit, is no stranger to scale. Upbit dominates the Korean market with an estimated 75-80% share. Its custody arm, Upbit Custody, holds a licensed VASP registration and offers a tech stack that includes 100% offline cold wallets, multi-party computation (MPC), distributed key generation (DKG), and multi-signature authentication. The service is marketed as a "real-time response regulatory infrastructure"—a phrase that deserves unpacking.
Core: The Architecture of Seized Assets
Let me walk through the technical layers, because the details matter more than the headlines.
Upbit Custody's solution for the police uses a three-tier security model. The first layer is physical isolation: all assets are stored in cold wallets with no direct internet connection. This is the industry standard for long-term storage of high-value assets, especially when the holdings are unlikely to be moved frequently. The second layer is key management: MPC and DKG distribute the private key into fragments, so no single party—neither Dunamu nor the police—can independently authorize a transaction. The third layer is transaction authorization: multi-signature requires multiple approvals before any asset movement.
This is not revolutionary. Fireblocks and BitGo have offered similar configurations for years. But the deployment context is what makes it interesting. The police need a "real-time response" capability—meaning they can request a freeze, unfreeze, or transfer of seized assets quickly when authorized by a court order. Balancing this with the security of an offline wallet creates a tension: every time a transaction must be executed, the cold wallet must be briefly connected to a signing environment. That moment—the "cold-to-warm" transition—is the most vulnerable point in the entire system.
Based on my audit experience with institutional custody setups, I estimate that Dunamu likely uses a hybrid approach: a small percentage of assets in a warm wallet for immediate response, with the bulk in deep cold storage. The article claims 100% cold storage, but that may refer to the static holdings before any operation. The real innovation here is not the technology—it's the service-level agreement with the police. The ability to respond on demand, with full audit trails, while maintaining MPC and multi-sig, is a governance challenge, not a technical one.
We audit the code, but who audits the conscience?
This brings me to the core insight: the police are outsourcing a slice of their sovereign power to a private company. When a law enforcement agency hands over custody of seized assets, it also hands over the operational responsibility for protecting those assets from external threats and internal abuse. The multi-sig model ensures that Dunamu cannot unilaterally move funds, but it also means that a police officer with a compromised key could collude with a Dunamu employee. The risk of insider collusion is low, but the impact would be catastrophic.
Contrarian: The Illusion of Decentralization in State Custody
Now, let me challenge the narrative. Many in the crypto community will celebrate this as a sign of institutional adoption—government trust in custody services. But I see a different signal. This arrangement centralizes trust in a single entity: Dunamu. If Dunamu's internal systems are compromised, or if a rogue employee with access to a key fragment is bribed, the entire stock of seized assets could be at risk. The police mitigate this with multi-sig, but they are still reliant on a single custodian. That is a single point of failure.
Moreover, the "real-time response" capability raises civil liberties questions. In a system where the police can request a freeze with minimal friction, what safeguards exist for erroneous seizures? The article does not mention an appeals process or a mechanism for asset owners to challenge a freeze. The law may provide for judicial oversight, but the operational speed of the system could outpace due process.
Build not for the peak, but for the plain.
This is not a criticism of Dunamu or the police. It is a reminder that infrastructure built for efficiency often sacrifices resilience. The plain, steady state of a well-functioning judiciary and transparent custody operations is what we should aim for, not the peak of real-time response that bypasses checks.
Takeaway: The Model That Will Be Copied
Despite these risks, the Korean model is likely to be replicated. Other countries facing similar challenges—the US, Singapore, Japan—are watching. The combination of a licensed custodian, MPC security, and government contract creates a template for how law enforcement can manage digital assets without building their own custody systems from scratch.
But the long-term question is not about technology. It is about accountability. Who monitors the monitor? The contract is only for one year. If Dunamu performs well, the relationship will deepen. If a scandal occurs, it will set the industry back. The strongest walls are not made of code, but of trust. And trust, like a cold wallet, is only as strong as the process that secures it.
Transparency is the new gold.
If Dunamu wants to lead, it must publish regular audits of its custody operations, including the number of assets held, the number of freeze requests, and the outcomes of those requests. Without transparency, this arrangement risks becoming a black box where the state's power meets private profit.
We audit the code, but who audits the conscience? The answer, for now, must be the public. And the public must demand more than a press release. They must demand the plain, boring details of how the system actually works.