On 21 February 2025, attackers drained roughly $1.5 billion in Ether from one of the world's largest crypto exchanges. They did not brute-force a password or steal a laptop from a vault. They tricked authorised signers into approving a transaction that looked legitimate on screen, from a multisignature cold wallet that was supposed to be the safest place of all.
That single event forced the industry to confront an uncomfortable truth: hot wallets, cold wallets and even multisig are not enough on their own. As tokenized assets and digital currencies become part of mainstream finance, custody has to evolve from protecting keys to protecting the assets themselves.
Custody basics: hot, cold and in between
Digital assets are controlled by private keys. Whoever can produce a valid signature with the key can move the asset. Custody is the discipline of generating, storing and using those keys safely.
Hot wallets
Hot wallets are connected to networks and systems, which makes them fast and convenient for trading, payments and daily operations. That same connectivity makes them the most exposed to attack.
Cold wallets
Cold wallets keep signing keys offline, often in air-gapped devices within secure facilities. They are much harder to reach remotely, which is why institutions use them for long-term reserves.
Warm and operational layers
Most institutions use a tiered model: a small hot float for liquidity, a larger cold reserve, and processes for moving funds between them.
The weakness of this model is not in the categories. It is in the moments when assets move, and in the assumption that whoever holds the key must be authorised.
What recent hacks actually teach us
Bybit, February 2025
Investigations by security researchers found that the attack was a supply chain compromise of the multisig wallet interface Bybit used. A developer workstation at the wallet provider was compromised earlier that month. When Bybit's signers approved a routine transfer from their Ether cold wallet, the interface displayed what looked like the correct transaction while the underlying smart contract logic had been altered. The signers unknowingly authorised a transaction that handed control to the attackers. The theft was attributed to North Korea's Lazarus Group.
Lesson: multiple signatures do not help if every signer is shown the same manipulated information. Security must verify what a transaction actually does, not just who approved it.
WazirX, July 2024
India's WazirX reported that about $234.9 million in digital assets was stolen from a multisig wallet on 18 July 2024. The wallet reportedly required three WazirX signatures and one signature from its custody provider. Attackers manipulated what signers saw and altered the wallet's smart contract, gaining control and draining funds. The attack was also linked to the Lazarus Group.
Lesson: the same pattern. Keys were not simply stolen; the approval process itself was subverted.
The bigger picture
Chainalysis reported that more than $3.4 billion in cryptocurrency was stolen in 2025, with losses concentrated in a small number of very large breaches. North Korea-linked hackers alone accounted for about $2.02 billion.
When the largest attacks dwarf the typical incident, the priority must be eliminating single catastrophic failure points.
The core problem: dumb assets
In traditional digital asset custody, the asset is "dumb." It obeys whoever produces a valid signature. If an attacker obtains the key, or tricks the key holders into signing, the asset moves instantly and irreversibly.
Programmable custody flips that model. It embeds security logic into the asset and its transfer rules, so even a valid signature cannot move the asset outside approved conditions.
Layer 1: stronger key management
Better key management remains the foundation.
Multi-Party Computation (MPC)
With MPC, a private key is never assembled in one place. Instead, key shares are held by multiple parties or devices, and they collaboratively produce a signature without any single party ever seeing the full key. Compromising one server or one employee is not enough.
The importance of this approach is reflected in standards work. The US National Institute of Standards and Technology (NIST) has published IR 8214C, its First Call for Multi-Party Threshold Schemes, inviting submissions for threshold cryptography that distributes trust over the creation, storage and use of keys.
Shamir's Secret Sharing
Keys, especially for deep cold storage and recovery, can be split into shards distributed across geographically separate vaults. A defined threshold of shards is needed to reconstruct or use them.
Hardware Security Modules (HSMs)
HSMs are tamper-resistant devices for generating and protecting keys. Institutions commonly look for validation under the US FIPS 140 standard. Note a timely detail: according to NIST's Cryptographic Module Validation Program, FIPS 140-2 validations move to the Historical List on 21 September 2026, making FIPS 140-3 validation the forward-looking benchmark for new deployments.
Layer 2: securing the approval process
The Bybit and WazirX incidents show that signing ceremonies need their own defences:
- Independent transaction verification, where signers confirm transaction contents on separate, trusted devices rather than one shared interface.
- Clear signing, where the actual contract call and destination are decoded and shown in human-readable form.
- Policy engines, which check every transaction against predefined rules before any signature is requested.
- Segregation of duties, so no one team controls both software deployment and signing.
Layer 3: programmable assets that protect themselves
This is where custody moves beyond hot and cold.
Native whitelisting at the asset layer
The asset can only be transferred to pre-approved, verified addresses. If an attacker obtains signing authority and tries to send funds to their own wallet, the transfer is rejected by the asset's own logic.
Time-locked and delayed execution
High-value outbound transfers from cold storage can be locked for a mandatory cooling-off period, such as 24 to 72 hours. That window gives security teams time to detect anomalies and freeze assets before a breach becomes final.
Threshold approval tied to the asset
Instead of relying only on a wallet's multisig configuration, approval thresholds, such as three of five board members, can be embedded in the token's transfer rules so they apply wherever the asset goes.
Context-aware conditions
Transfers can require contextual signals, such as requests originating from approved networks, during business hours, or with biometric confirmation from authorised executives, verified through trusted data sources.
Kill switches and clawbacks
Authorised administrators, under strict governance, can trigger a freeze or recovery function if a breach is detected, returning assets to a secure treasury contract.
Balancing security with usability
Programmable controls introduce trade-offs that must be managed thoughtfully.
- Delays reduce agility. Time-locks should apply to high-value or unusual transfers, not routine operational flows.
- Admin powers must be governed. A freeze function is itself a target. It needs multi-party authorisation, logging and oversight.
- Whitelists need maintenance. Adding counterparties must follow its own secure, time-delayed process.
- Interoperability matters. Rules must hold when assets move across networks, rather than being lost at a bridge.
Who needs programmable custody
- Central banks and CBDC issuers, protecting national digital currency reserves with multi-party controls that prevent unauthorised issuance or movement.
- Commercial banks and asset managers, holding tokenized bonds and real estate with rules that allow transfers only to KYC-approved counterparties.
- Exchanges and trading firms, protecting customer funds in cold storage with time-locks and whitelisting.
- Corporate treasuries, holding digital assets under board-level approval requirements.
- Stablecoin and RWA issuers, who must prove reserves and prevent unauthorised minting.
A practical custody checklist
- Eliminate single points of failure with MPC, threshold signatures and distributed key shards.
- Verify what is signed, not just who signs, using independent devices and clear signing.
- Enforce policies before signatures, with automated rules on amounts, destinations and timing.
- Add asset-level controls such as whitelisting and time-locks for high-value holdings.
- Plan for breach response, with governed freeze and recovery functions.
- Monitor continuously, with anomaly detection on every outbound request.
- Audit the whole stack, including third-party wallet software and cross-chain messaging layers.
- Prove reserves, reconciling on-chain supply against off-chain backing.
Questions boards should ask about custody
Custody failures are not just technical incidents. They are governance failures with balance sheet consequences. Boards and risk committees overseeing digital assets should be able to get clear answers to these questions.
- Where is our single point of failure? Is there any one person, device, vendor or interface whose compromise would allow assets to move?
- How do signers know what they are signing? Are transactions independently verified outside the primary interface?
- What limits apply automatically? Are there enforced caps, whitelists and delays on high-value transfers?
- How fast can we freeze assets? Who can trigger it, how is that power controlled, and has the process been tested?
- Which third parties can influence our wallets? How are their software updates reviewed and monitored?
- How are reserves proven to customers and auditors?
- When was the last independent security audit, and what did it find?
If management cannot answer confidently, the organisation is relying on luck rather than design.
How E1 envisions programmable custody
E1 is building next-generation digital asset custody designed to move beyond static private keys, with the security logic embedded in the asset itself.
E1's approach includes:
- Institutional hot wallet infrastructure designed for low-latency signing and automated liquidity rebalancing that sweeps excess funds to cold storage based on real-time risk parameters.
- Deep cold storage architecture designed around air-gapped, geographically distributed vaults and Shamir's Secret Sharing.
- Advanced key management using Multi-Party Computation and integration with certified Hardware Security Modules.
- Programmable digital assets designed with native whitelisting, 24 to 72 hour time-locks for high-value outbound transfers, asset-level threshold approvals, context-aware conditions, and governed kill switches and clawbacks.
- Cross-chain custody powered by Datachain, designed to move assets between networks without exposing them to vulnerable third-party bridge contracts.
- Proof of Reserves, designed to continuously reconcile on-chain supply against off-chain fiat balances connected through Chainpay.
The hacks of 2024 and 2025 showed that the old model, where a wallet simply obeys whoever holds the key, is too fragile for the future of finance. E1's vision is custody where assets defend themselves, so that even a stolen key or a manipulated signing screen is not enough to take them.
Sources
