How can I verify this generator is truly zero-transmission?

Open browser DevTools (F12), go to the Network tab, clear existing entries, then generate a password. You will see zero network requests. The live audit panel on the right of the tool also confirms this in real time. The password is generated entirely by crypto.getRandomValues() in your browser — nothing is transmitted to any server.

What is the difference between the threat model presets?

Standard (20 characters) is appropriate for general consumer accounts. Professional (24 characters) suits business credentials and accounts that may be subject to targeted attacks. Journalist (32 characters) is for high-threat users including journalists, activists, and others facing state-level adversaries. Air-Gap (48 characters) is for credentials used on air-gapped systems or the highest-threat environments.

What entropy source does this generator use?

All generation uses crypto.getRandomValues() from the W3C Web Cryptography API — the browser's implementation of the operating system CSPRNG. On Linux this is /dev/urandom seeded from hardware entropy. On macOS it is Fortuna. On Windows it is CNG. This is identical to the entropy source used by security-critical applications and HSMs.

Is this generator appropriate for generating PGP passphrases?

Yes, with one caveat: PGP passphrases typically need to be memorised rather than stored in a password manager. The random output from this generator can be used to seed a Diceware word selection — generate random numbers and use them to index into the EFF Large Wordlist. See our guide on PGP passphrase selection for the full process. Verify zero-transmission via DevTools before generating any credential for high-sensitivity use.

Why does the Journalist preset use 32 characters instead of 20?

For high-threat users, 32 characters provides a larger safety margin against future advances in computing. At 20 characters the password is already computationally infeasible to crack — but 32 characters ensures this remains true even if significant improvements in attacking hardware occur. The additional length costs nothing in a generated credential and provides a meaningful buffer.

What does the Privacy Audit panel show?

The Privacy Audit panel shows in real time: the number of network requests made during your session, bytes transmitted, localStorage keys present, and the entropy source used. It updates live as you interact with the tool. A fully zero-transmission session shows zero requests and zero bytes transmitted.

Can I use this tool on Tor Browser?

Yes. The generator is designed to work on Tor Browser. JavaScript must be enabled (Security Level Standard or Safer). The tool makes no requests that would reveal your identity — it loads once and then operates entirely client-side. For maximum privacy, load the page, disconnect from the network, generate your credential, then copy it before reconnecting.

What is the NCSC guidance on password generation?

The UK National Cyber Security Centre recommends machine-generated random passwords stored in a password manager over user-chosen passwords. NCSC guidance emphasises length over complexity — longer random passwords are stronger than shorter complex ones. The generator defaults align with NCSC recommendations: 20+ characters, CSPRNG, all character classes.

Is this suitable for generating credentials for SecureDrop?

SecureDrop account credentials can be generated here, but they should be memorised rather than stored in a cloud-synced manager. Use the Journalist preset (32 characters) or Air-Gap preset (48 characters) and memorise the credential using spaced repetition or a memory palace. Verify zero-transmission via the Network tab before generating. Ideally, access this page over Tor on a dedicated device.

What compliance standards does this tool meet?

The generator uses crypto.getRandomValues() which meets NIST SP 800-90A requirements for approved DRBG seeding. Generated passwords at 20+ characters exceed NIST SP 800-63B 2025 minimum length requirements and NCSC guidance. The privacy-by-design architecture aligns with CISA Secure by Design principles and ICO guidance on data minimisation under UK GDPR.

Secure Key Generator — Privacy-First Zero-Transmission Password Generator

Open DevTools (F12), go to Network, clear entries, generate a password. You will see zero requests. The live audit panel also confirms this in real time. The password is generated by crypto.getRandomValues() in your browser only.

Standard (20 chars) for general accounts. Professional (24 chars) for business credentials and targeted attack risk. Journalist (32 chars) for high-threat users including journalists and activists. Air-Gap (48 chars) for credentials used on air-gapped systems or the highest-threat environments.

crypto.getRandomValues() from the W3C Web Cryptography API — backed by the OS CSPRNG. On Linux: /dev/urandom. On macOS: Fortuna. On Windows: CNG. This is the same entropy source used by security-critical applications and HSMs.

Yes. The generator works on Tor Browser at Security Level Standard and Safer. JavaScript must be enabled. The tool makes no requests that would reveal your identity — it loads once, then operates entirely client-side. For maximum privacy, load the page, disconnect from the network, generate your credential, then reconnect.

Yes — use the Journalist (32 chars) or Air-Gap (48 chars) preset. PGP passphrases should be memorised rather than stored in a cloud-synced manager. Verify zero-transmission via DevTools before generating. See our PGP passphrase guide →

Real-time counts of: network requests made, bytes transmitted, localStorage keys present, entropy source used, and cookie count. It updates live as you interact. A zero-transmission session shows 0 requests and 0 bytes transmitted at all times.

The NCSC recommends machine-generated random passwords stored in a password manager. The NCSC explicitly discourages mandatory periodic rotation without evidence of compromise. All presets align with NCSC length and randomness guidance. The tool's zero-transmission design goes beyond NCSC requirements by eliminating the server as a potential data point entirely.

crypto.getRandomValues() meets NIST SP 800-90A approved DRBG seeding requirements. 20+ character passwords exceed NIST SP 800-63B 2025 minimums and NCSC guidance. The privacy-by-design architecture aligns with CISA Secure by Design principles and UK ICO data minimisation guidance under UK GDPR.

For high-threat users, 32 characters provides a larger margin against future computing advances. At 20 characters the password is already computationally infeasible to crack — 32 ensures this remains true even with significant hardware improvements. The additional length costs nothing in a generated credential.

No. HTTPS encrypts transmission but the server still knows what was generated. The operator can log, sell, or be legally compelled to disclose generated passwords. Only client-side generation — verifiable via the Network tab — eliminates this risk. See our guide: zero-knowledge generation →

Privacy by verification, not by promise

Every security claim on this site can be independently verified in your browser. No trust required.

🔬

Live network audit

The audit panel shows real-time request counts and bytes transmitted. Open DevTools to cross-reference — both will show zero during generation.

🎯

Threat model presets

Four presets from Standard (20 chars) to Air-Gap (48 chars), calibrated to realistic attack scenarios for each audience tier.

🔒

CSPRNG only

Exclusively uses crypto.getRandomValues() — the W3C Web Cryptography API backed by OS hardware entropy. Never Math.random().

🌍

Works on Tor Browser

Designed to function at Security Level Standard and Safer on Tor Browser. JavaScript required — no WebRTC or canvas fingerprinting.

Compliance and standards reference

All presets exceed the requirements of every listed framework. Sources: NIST, NCSC, OWASP, CISA, EFF.

Framework / Guidance	Min Length	Entropy Source	Rotation	SKG Compliance
NIST SP 800-63B 2025	15 chars	CSPRNG required	SHALL NOT (prohibited)	✓ Standard (20+)
NIST SP 800-90A	—	Approved DRBG	—	✓ crypto.getRandomValues()
NCSC Password Guidance	No minimum	Machine-generated preferred	On compromise only	✓ All presets
OWASP Cryptographic Storage	—	OS CSPRNG required	—	✓ OS entropy source
CISA Secure by Design	—	Secure defaults	—	✓ Zero-transmission default
EFF Surveillance Self-Defence	6+ Diceware words	Physical or CSPRNG	On compromise	✓ Journalist (32 chars ≈ 7+ words)
Freedom of the Press Foundation	Strong random	CSPRNG or physical dice	On compromise	✓ All presets with verification

Why privacy-first generation matters

81%

of breaches involve weak or stolen credentials

Verizon DBIR 2025

14B+

compromised passwords in breach databases

HaveIBeenPwned 2026

network requests made during generation — verifiable in DevTools

securekeygenerator.com

131

bits entropy — Standard preset (20 chars, full ASCII pool)

H = L × log₂(N)

Written by privacy researchers

The research and tools on this site are written by AY Tanoli, a privacy researcher and security consultant who has trained journalists, activists, and human rights workers in operational security across Europe, Southeast Asia, and the MENA region. Dr. Chen's work draws from EFF Surveillance Self-Defence, Freedom of the Press Foundation training materials, and NCSC guidance.

All technical claims are sourced from primary documents. The tool itself can be verified without trusting any claim — open DevTools and watch the network activity during generation.

🗝️ KeePassXC (Local Only)

📡 Bitwarden (Self-Hosted)

🛡️ YubiKey (Hardware 2FA)

The only generator that proves nothing
was transmitted.

Zero Transmission Proof Generator

🔐 securekeygenerator.com — generator

// privacy audit — live

Privacy by verification, not by promise

Live network audit

Threat model presets

CSPRNG only

Works on Tor Browser

Compliance and standards reference

Why privacy-first generation matters

Privacy-respecting security tools

Written by privacy researchers

// portfolio

Frequently asked questions

Privacy & security research