Crypto exchanges should not rely on cold-storage percentages alone to judge security, according to cybersecurity firm CertiK. The company argues that custody strategy must be evaluated through measurable limits on how much value remains in hot wallets—the part of exchange infrastructure most exposed to hacking.

The call comes after a wave of public disclosures from exchanges about cold-wallet holdings following the record Binance hack, which targeted the platform’s hot wallets. Gate reported that 88.26% of its total assets were held in cold storage, while Coinbase said it keeps more than 99% of customer assets offline. CertiK says those ratios can be informative, but they should not be treated as a complete security score.

Key takeaways

CertiK says cold-storage ratios reflect online exposure, but do not capture the full security picture.
Exchanges should use hard dollar caps on hot-wallet balances—potentially per asset and per blockchain—rather than only percentage targets.
CertiK warns that rigid ratios can complicate operations, including withdrawals and liquidity management.
The firm proposes a framework combining multiple limits: customer-liability coverage, estimated withdrawal needs, and an absolute loss ceiling tied to the exchange’s recovery capacity.

Why cold-storage ratios aren’t enough

CertiK’s research team told Cointelegraph that cold storage remains a useful indicator of online exposure, but it cannot stand alone as a security metric. The reason is straightforward: what matters operationally is not just how much is stored offline, but how much value sits within systems that must stay online to support trading and withdrawals.

Cold-storage frameworks are increasingly referenced by regulators and industry watchers. CertiK highlighted several jurisdictional requirements: South Korea mandates at least 80% of customers’ crypto assets by value in cold storage, Singapore calls for at least 90%, Japan requires at least 95%, and Hong Kong’s baseline for licensed trading platforms is 98%. Those rules are designed to reduce the amount of customer funds exposed to direct online risks.

However, CertiK notes that the risk reduction from moving higher and higher into cold storage may taper off at very high levels. In other words, after a certain point, chasing incremental percentage improvements may yield diminishing returns compared with improving how hot-wallet risk is bounded and managed.

The operational trade-offs regulators don’t measure

CertiK also points to practical drawbacks of enforcing strict storage ratios without flexibility. If an exchange is forced to keep a fixed cold-storage percentage regardless of day-to-day conditions, it may struggle to respond quickly when customers want to withdraw.

According to CertiK, rigid approaches can lead to forced liquidity moves—such as rushed transfers to meet operational demands—while also potentially increasing pressure on key-management systems. Those systems are central to controlling access to funds and can become overloaded if operational liquidity planning isn’t built around realistic withdrawal behavior and execution timelines.

For traders and users, these issues matter because custody policy ultimately intersects with service reliability. A security plan that slows withdrawals or creates unpredictable execution risk could degrade the user experience at exactly the moment when customers are most likely to seek access to funds.

CertiK’s proposed framework: hot-wallet caps and circuit breakers

CertiK’s alternative centers on setting hot-wallet limits in hard dollar terms rather than relying only on offline storage ratios. The firm recommends that exchanges cap hot-wallet holdings both by percentage and by absolute exposure, and that these limits should be calculated per asset and blockchain.

Specifically, CertiK says exchanges should consider the strictest of three boundaries:

A percentage of customer liabilities (so hot-wallet exposure never exceeds what the exchange owes, as a proportionate constraint).
Estimated withdrawal needs (to ensure the exchange can meet customer demand without unsafe overexposure).
An absolute loss ceiling based on the exchange’s financial resources and recovery capabilities (so that even a worst-case hot-wallet incident is capped at a survivable level).

The core idea is to balance liquidity and safety. CertiK frames this as a structure where flexibility supports day-to-day operations, while hard caps and “circuit breakers” limit how much an attacker could plausibly extract before controls stop or constrain further loss.

Binance updates and the broader aftermath

While CertiK’s guidance focuses on how exchanges should design custody risk controls, recent events show why hot-wallet exposure remains a sensitive topic. Last week, Binance said it had replenished approximately 98% of affected tokens after the hack and would absorb the losses. In a later recovery update, the exchange said remaining tokens were less liquid and would take longer to acquire.

Investigations into the laundering attempts after the breach also continued. According to Elliptic, North Korea-linked hackers created memecoins intended to help launder stolen funds from the incident. Elliptic said it traced roughly $430,000 through six newly launched tokens on NEAR, using liquidity pools to obscure transfers between wallets, while noting that the activities remained traceable on-chain.

For market participants, the combination of rapid “replenishment” messaging and ongoing follow-on laundering efforts highlights the multi-stage nature of major exchange incidents. Custody security policy is therefore not just a question of offline versus online storage, but of how quickly and safely an exchange can restore operational funds while limiting the attacker’s ability to translate stolen value into durable profit.

Going forward, investors and users should watch whether exchanges move beyond cold-storage headline percentages toward transparent, enforceable hot-wallet exposure limits—especially caps that reflect withdrawal demand and financial recovery capacity, not just regulatory minimums.