Bitcoin Backup Encryption: BIP-38 vs SLIP-39 vs Codex32
Compare Bitcoin backup encryption methods: BIP-38, BIP-39 passphrase, SLIP-39 Shamir, Codex32, and SeedXOR by security, wallet support, and recovery.
Bitcoin Backup Encryption Methods Compared
A Bitcoin backup is only as strong as its weakest link. Writing down a seed phrase on paper protects against device failure, but it creates a new risk: anyone who finds that paper controls your funds. Encryption and secret splitting address this by adding a layer of protection to the backup itself.
Five methods dominate the landscape: BIP-38 encrypted private keys, BIP-39 passphrases (the "25th word"), SLIP-39 Shamir backup, Codex32 (BIP-93) hand-verifiable shares, and SeedXOR. Each makes different tradeoffs between security strength, hardware wallet compatibility, recovery complexity, and human verifiability.
| Method | Standard | Type | Threshold | Max Shares | Hand-Verifiable | Key Wallet Support |
|---|---|---|---|---|---|---|
| BIP-38 | BIP-38 (2012) | Passphrase encryption | N/A (single key) | N/A | No | Limited (legacy) |
| BIP-39 Passphrase | BIP-39 (2013) | Seed derivation modifier | N/A | N/A | No | Nearly universal |
| SLIP-39 | SLIP-39 (2017) | Shamir secret sharing | k-of-n | 16 per group | No | Trezor, Keystone |
| Codex32 | BIP-93 (Draft) | Shamir + checksum | k-of-n (k: 2-9) | 31 | Yes | Coldcard, Sparrow |
| SeedXOR | Coinkite (2020) | XOR splitting | n-of-n only | Unlimited | Yes (manual XOR) | Coldcard |
For a broader look at physical backup media including metal plates and paper, see our Bitcoin seed storage comparison.
BIP-38: Passphrase-Encrypted Private Keys
BIP-38, proposed by Mike Caldwell and Aaron Voisine in 2012, encrypts individual private keys under a passphrase. The output is a 58-character Base58Check-encoded string that can be printed on paper wallets or stamped into metal. The scheme uses scrypt for key derivation (N=16384, r=8, p=8) and AES-256 for encryption.
BIP-38 was designed for the era of single-key paper wallets. It encrypts one private key at a time, not an HD wallet seed. This means each address requires its own encrypted backup. The BIP's comment summary on the Bitcoin BIPs repository reads "Unanimously Discourage for implementation," reflecting the community's shift toward HD wallets and seed-based backups.
BIP-38 remains relevant for one use case: physical bearer instruments like Casascius coins and OpenDime devices, where a single encrypted key needs to survive years of storage. For modern HD wallets, the other methods in this comparison are more appropriate.
BIP-39 Passphrase: The 25th Word
The BIP-39 passphrase (often called the "25th word") is an optional string appended to the seed phrase during key derivation. It is passed alongside the mnemonic into PBKDF2 with 2,048 rounds of HMAC-SHA512 to produce the master seed. When no passphrase is set, an empty string is used by default.
Any non-empty passphrase produces an entirely different wallet with different keys and addresses. This creates a natural duress feature: a user can keep a small balance in the passphrase-less wallet and store the real funds behind a passphrase. The critical risk is that a mistyped passphrase does not produce an error. It silently opens an empty wallet, making it easy to believe funds are lost when the passphrase was simply entered incorrectly.
Hardware wallet support is nearly universal. Trezor, Ledger, Coldcard, BitBox02, Foundation Passport, Keystone, and Blockstream Jade all support BIP-39 passphrases. Some devices store the passphrase temporarily in memory and require re-entry each session, while others allow saving it behind a secondary PIN.
The passphrase itself becomes a single point of failure. If forgotten, funds cannot be recovered even with the correct seed phrase. Best practice is to store the passphrase separately from the seed, ideally in a different physical location, and to test recovery on a fresh device before depositing significant funds. For more on seed entropy and passphrase strength, see our research on seed phrase entropy and security.
SLIP-39: Shamir Backup
SLIP-39, developed by SatoshiLabs, implements Shamir's Secret Sharing for wallet backups. Instead of a single seed phrase, the master secret is split into multiple 20-word shares from a dedicated word list (not the same as the BIP-39 word list). A configurable threshold determines how many shares are needed to reconstruct the secret: for example, a 3-of-5 scheme means any three shares can recover the wallet, but fewer than three reveal nothing.
SLIP-39 supports groups of shares, enabling complex access structures. A corporate treasury might require 2-of-3 executive shares AND 2-of-3 security team shares. Each group has its own threshold, and an overall group threshold determines how many groups must participate.
The tradeoff is compatibility. SLIP-39 uses a different word list and a different derivation path than BIP-39. It is not possible to convert a SLIP-39 backup into a BIP-39 mnemonic or vice versa without third-party tools. Native hardware wallet support is limited to Trezor (which made it the default on its Safe Family devices as of June 2024) and Keystone. Software wallets including Sparrow, Electrum, and BlueWallet have added SLIP-39 import support. Ledger, Coldcard, BitBox02, and Foundation Passport do not support SLIP-39 natively.
Codex32: Hand-Verifiable Shamir Shares
Codex32, specified in draft BIP-93, takes Shamir secret sharing a step further. Authored by Leon Olsson Curr, Pearlwort Sneed, and Andrew Poelstra, it encodes BIP-32 master seeds as Bech32 strings with a BCH error-correcting checksum. The checksum can detect up to 8 character errors and locate up to 4 errors in a single share.
The defining feature is hand-verifiability. Generation, splitting, checksum verification, and secret reconstruction can all be performed without electronic computers. Blockstream publishes tear-out worksheets and paper "volvelles" (rotating disc calculators) for this purpose. For users who distrust every electronic device, Codex32 is the only backup method that eliminates computer trust entirely.
Codex32 supports up to 31 shares with thresholds between 2 and 9. Shares use the lowercase bech32 alphabet with no mixed case, which makes them well suited for stamping into metal plates. Coldcard added Codex32 support in firmware v5.6.3 (Mk4/Mk5) and v1.5.3Q (Q series), allowing on-device generation, import, splitting, and share derivation. Sparrow Wallet also supports Codex32 import.
SeedXOR: XOR-Based Seed Splitting
SeedXOR, introduced by Coinkite in Coldcard firmware v4.1.0, splits a BIP-39 seed phrase into two or more parts by XOR-ing the underlying entropy. Each part is itself a valid BIP-39 phrase with a correct checksum, which means each part can be loaded into any wallet as a functional (but different) wallet.
This creates plausible deniability: if one part is discovered, it appears to be a normal seed phrase and can even hold a small "honeypot" balance. An attacker cannot determine whether a seed is a standalone wallet or a SeedXOR component without possessing all other parts.
The limitation is that SeedXOR requires all parts for recovery (n-of-n only). There is no threshold scheme: losing any single part permanently destroys access to the original seed. The XOR operation can be performed by hand on paper using Coinkite's published worksheets, though it is more tedious than Codex32's volvelle method. SeedXOR is primarily a Coldcard feature, and no other major hardware wallet vendor has adopted it.
Hardware Wallet Compatibility
Wallet support is often the deciding factor when choosing a backup method. The following table shows current compatibility across major hardware wallets.
| Hardware Wallet | BIP-38 | BIP-39 Passphrase | SLIP-39 | Codex32 | SeedXOR |
|---|---|---|---|---|---|
| Trezor Safe 3 / Safe 5 | No | Yes | Yes (native) | No | No |
| Ledger Nano S Plus / X | No | Yes | No | No | No |
| Coldcard Mk4 / Q | No | Yes | No | Yes | Yes |
| BitBox02 | No | Yes | No | No | No |
| Foundation Passport | No | Yes | No | No | No |
| Keystone 3 Pro | No | Yes | Yes | No | No |
| Blockstream Jade | No | Yes | No | No | No |
BIP-39 passphrase is the only method with universal hardware support. SLIP-39 is locked into the Trezor and Keystone ecosystem. Codex32 and SeedXOR are Coldcard features. This fragmentation means your choice of backup method can constrain which hardware you can use for recovery: an important consideration for inheritance planning, where the person recovering may not own the same device.
Metal Backup Compatibility
Physical durability matters for long-term storage. Each backup format interacts differently with metal stamping plates and engraving tools.
- BIP-39 phrases use the standard 2,048-word list. Most metal backup products (Cryptosteel, Billfodl, Blockplate) are designed around this format, typically storing the first 4 letters of each word.
- SLIP-39 shares use a different 1,024-word list with 20 words per share. Standard BIP-39 metal plates work if they support custom words, but dedicated SLIP-39 plates are uncommon.
- Codex32 uses the lowercase bech32 alphabet (32 characters, no mixed case). This makes it ideal for stamping: fewer characters, no ambiguity between upper and lowercase. Standard letter-punch sets work well.
- SeedXOR parts are valid BIP-39 phrases, so any BIP-39 metal backup product works directly.
- BIP-38 strings are 58-character Base58Check strings. They can be stamped, but the length and mixed-case format make errors more likely.
For users planning metal backups with Shamir splitting, Codex32's bech32 encoding is the most stamp-friendly format. SLIP-39 shares require more space (20 words each) but benefit from the same error resilience as standard word-based backups: even with minor transcription errors, the words remain recognizable.
Security Comparison
Security properties vary substantially across these methods. The table below compares encryption strength, brute-force resistance, and failure modes.
| Property | BIP-38 | BIP-39 Passphrase | SLIP-39 | Codex32 | SeedXOR |
|---|---|---|---|---|---|
| Encryption | scrypt + AES-256 | PBKDF2 (2048 rounds) | None (splitting only) | None (splitting only) | None (splitting only) |
| Brute-force resistance | High (scrypt-hard) | Moderate (PBKDF2) | Information-theoretic | Information-theoretic | Information-theoretic |
| Error detection | 4-byte checksum | None (silent failure) | Checksum per share | BCH code (8 errors) | BIP-39 checksum |
| Single point of failure | Passphrase | Passphrase | No (threshold) | No (threshold) | Yes (all parts needed) |
| Duress/decoy wallet | No | Yes | No | No | Yes (each part is a wallet) |
Shamir-based methods (SLIP-39 and Codex32) provide information-theoretic security: possessing fewer than the threshold number of shares reveals zero information about the secret, regardless of computational power. This is fundamentally stronger than passphrase-based encryption, where a weak passphrase can be brute-forced. However, Shamir methods require distributing and protecting multiple physical shares, which introduces operational complexity.
How to Choose a Backup Method
The right method depends on your threat model, your hardware wallet ecosystem, and how many trusted locations you have for storing backup material.
- Use a BIP-39 passphrase if you want maximum wallet compatibility and can commit a strong passphrase to memory or store it separately from your seed phrase. This is the simplest upgrade from an unprotected seed.
- Use SLIP-39 if you are in the Trezor or Keystone ecosystem and want threshold-based recovery without trusting a single backup location. Ideal for families or small teams distributing shares to trusted parties.
- Use Codex32 if you prioritize verifiability and want to confirm your backup integrity without powering on any device. Best for long-term cold storage where you may not check the backup for years.
- Use SeedXOR if you want plausible deniability (each part looks like a normal wallet) and are comfortable with n-of-n recovery. Good for personal use with two or three storage locations.
- Avoid BIP-38 for new setups. It encrypts individual keys rather than HD seeds and has limited modern wallet support.
For users holding Bitcoin on Spark, seed backup practices apply just as they do for any self-custodial Bitcoin wallet. The self-custody model means your backup is your only recovery path: choosing the right encryption method is a one-time decision with permanent consequences.
You can also combine methods. A common pattern is to use a BIP-39 passphrase for day-to-day protection and Codex32 or SLIP-39 for the seed phrase itself. The BIP-85 standard can derive child seeds from a master seed, allowing you to use different backup strategies for different wallets derived from the same root.
Frequently Asked Questions
Should I use Shamir splitting or a passphrase?
It depends on your operational setup. A passphrase is simpler: you protect one seed phrase and memorize (or separately store) one passphrase. Shamir splitting (SLIP-39 or Codex32) eliminates the single point of failure by distributing shares, but it requires multiple secure storage locations and compatible recovery hardware. If you have three or more trusted locations (a home safe, a bank vault, a relative's house), Shamir splitting with a 2-of-3 threshold provides better resilience. If you have only one or two storage locations, a strong passphrase is more practical.
Can I verify my Codex32 shares without a computer?
Yes. Codex32 was specifically designed for hand verification. Using printed worksheets and paper volvelles published by the authors, you can compute and verify the BCH checksum of each share entirely by hand. This process takes roughly 30 minutes per share and can detect up to 8 transcription errors. No other backup method offers this level of offline verification.
Can I convert between SLIP-39 and BIP-39?
Not directly. SLIP-39 and BIP-39 use different word lists and different derivation algorithms. You cannot take a SLIP-39 share set and produce an equivalent BIP-39 mnemonic, or vice versa, without moving funds to a new wallet. Third-party tools like python-slip39 can back up an existing BIP-39 phrase as SLIP-39 share cards, but this is a backup layer on top of the original phrase rather than a format conversion. Recovery still requires SLIP-39-compatible software.
What happens if I lose one SLIP-39 share?
In a k-of-n scheme, losing a single share does not prevent recovery as long as you still hold at least k shares. For example, in a 3-of-5 setup, you can lose up to two shares and still recover your wallet. The lost share reveals no information about the secret on its own. However, losing shares reduces your margin of safety: if you start with 3-of-5 and lose two shares, you are down to exactly 3 remaining shares with no room for further loss.
Is SeedXOR as secure as Shamir secret sharing?
SeedXOR provides information-theoretic security (possessing n-1 parts reveals nothing), which matches Shamir's security guarantee. The difference is flexibility: Shamir allows k-of-n thresholds (e.g., 2-of-3), while SeedXOR always requires all parts. This makes SeedXOR less fault-tolerant. If you lose one SeedXOR part, your funds are permanently inaccessible. With a 2-of-3 Shamir scheme, you can lose one share and still recover.
Which method is best for Bitcoin inheritance planning?
For inheritance planning, prioritize wallet compatibility and clear recovery instructions. SLIP-39 with a 2-of-3 threshold works well when shares are distributed to a lawyer, a family member, and a safe deposit box. Codex32 is another strong option because its error-detecting checksum reduces the risk of transcription errors during a stressful recovery process. Avoid methods with limited wallet support (like SeedXOR) unless you include a compatible device with the backup materials. See our research on Bitcoin inheritance planning for detailed guidance.
Do Shamir shares need to be stored on metal?
Metal storage is recommended for any backup intended to survive decades. Paper degrades from moisture, fire, and UV exposure. For Shamir shares specifically, each share should be stored on its own metal plate in a separate location. Codex32 shares are particularly well suited to metal because the bech32 alphabet is lowercase-only and unambiguous. SLIP-39 shares can be stamped using any metal plate that supports custom word entry.
This tool is for informational purposes only and does not constitute financial advice. Data is approximate and based on publicly available specifications and vendor documentation as of late 2026. Hardware wallet support may change with firmware updates. Always verify compatibility with your specific device and practice a full recovery before relying on any backup method.
Build with Spark
Integrate bitcoin, Lightning, and stablecoins into your app with a few lines of code.
Read the docs →
