User-controlled opt-in to skip the master-password prompt on subsequent
app starts. The vault state (raw AES key + salt + username + session
token) is bundled and handed to the Delphi side, which DPAPI-encrypts
it with CRYPTPROTECT_CURRENT_USER and stashes the blob at
%LOCALAPPDATA%\PMServer\quickunlock.bin.
Honest threat model
===================
This is NOT biometric authentication. The DPAPI scope is the Windows
USER ACCOUNT — any process running as the same user can decrypt the
blob via the same DPAPI call. The security perimeter is the Windows
account itself. The Settings UI label is "Quick unlock" with an
explainer:
"convenient on a personal machine, not safe on a shared one"
If the user has Windows Hello / fingerprint / PIN configured at the
OS level, then Windows login is biometric-gated and that gating
transitively applies to DPAPI access — but the cryptographic strength
of the encryption isn't tied to the biometric, it's tied to the
Windows account secret. Honest framing matters here, so the feature
isn't sold as "biometric".
Backend
=======
New unit Source/PM.QuickUnlock.pas:
- StoreQuickUnlock(bytes) → DPAPI-encrypt and persist to
%LOCALAPPDATA%\PMServer\quickunlock.bin
- LoadQuickUnlock(out bytes) → read file, DPAPI-decrypt
- ClearQuickUnlock → forget-me
- HasQuickUnlock → file existence probe
DPAPI declarations are local (CryptProtectData / CryptUnprotectData
from crypt32.dll) — Winapi.WinCrypt's signatures drift across Delphi
versions and we don't want to fight that.
Bridge commands (UMainForm.HandleBridgeCommand):
cmd://quickunlock/store?data=<base64> payload opaque to Delphi
cmd://quickunlock/get → ExecuteJavaScript callback
Bridge.onQuickUnlockResult(b64|null)
cmd://quickunlock/clear forget-me
cmd://quickunlock/status → Bridge.onQuickUnlockStatus(bool)
The get / status results are returned via ExecuteJavaScript rather than
HTTP (the bridge is request-only) — JS resolves a Promise that the
caller awaited.
Client
======
state.quickUnlockEnabled mirrors localStorage flag, lazy-cleared if the
backing DPAPI blob has gone missing (e.g., user reset Windows profile).
enableQuickUnlock():
1. askReauth + /reauth to verify it's actually the user.
2. exportKey('raw', state.cryptoKey) — extractable already.
3. JSON-bundle { v, username, salt, token, csrf, key } → base64.
4. cmd://quickunlock/store sends the blob to Delphi.
tryQuickUnlock() (called from init):
1. Probe localStorage flag.
2. cmd://quickunlock/get, await Bridge.onQuickUnlockResult.
3. Decode JSON, importKey, restore state.* + sessionStorage.
4. Return true on success, false to fall through to master-pw login.
Two restore scenarios both covered:
A. Same app session (sessionStorage still populated, only cryptoKey
was wiped by lock). tryQuickUnlock just restores the key.
B. Cold start (sessionStorage empty). tryQuickUnlock restores
EVERYTHING from the DPAPI blob, including the session token.
UI
==
Settings panel → new "Quick unlock" section above Recovery key.
Single toggle button: "Enable on this device" / "Disable" with status
line above. Opens settings → bridgeQuickUnlockStatus() reconciles the
JS-side flag with the actual file (drift detection).
Stale-blob protection
=====================
The stored blob holds the AES key BYTES, which would become useless
if the vault were re-encrypted under a different key. Three paths
that re-encrypt the vault now also wipe the DPAPI blob:
- Explicit doLogout (user said "I'm done")
- Master password change (new key, old blob can't decrypt anything)
- (Recovery redeem already forces master pw change → covered.)
The blob persists across the passive lockVault() flow on purpose —
that's the whole point: lock without losing convenience.
Init wiring
===========
On app start, the existing "restore session from sessionStorage" path
now falls through to tryQuickUnlock if either sessionStorage is empty
OR the cryptoKey is gone. Auth screen shows up only after both
attempts fail.
In a zero-knowledge vault, forgetting the master password normally
means losing the data — the AES key is derived from the master pw
and the server can't help. This commit adds the standard escape
hatch: a one-time recovery code that key-wraps the AES key so the
user can get back in.
Threat model
============
The plaintext recovery code is shown to the user exactly once, at
generation time. Server only ever stores SHA-256(code) + an AES-GCM
wrap of the vault key under a KEK = PBKDF2(code, kdf_salt, 600k).
Without the plaintext code the server cannot unwrap. The code is
high-entropy (96 bits from a 32-char ambiguity-free alphabet, in 4
groups of 4) — printed form is misreading-resistant.
Single use: redeeming deletes the row inside the same DB.Lock the
lookup happened in, so concurrent redeem attempts are race-free.
Failed redemptions feed both the per-IP rate limit AND the per-
username lockout, so brute-forcing the code is infeasible.
Schema
======
recovery_keys (
user_id INTEGER PRIMARY KEY (1:1 with users, FK cascade),
code_hash TEXT NOT NULL (SHA-256 hex of plaintext code),
kdf_salt TEXT NOT NULL (PBKDF2 salt for KEK derivation),
wrapped_key TEXT NOT NULL (base64 AES-GCM ciphertext of vault key),
wrapped_iv TEXT NOT NULL (base64 12B IV for the wrap),
created_at DATETIME DEFAULT CURRENT_TIMESTAMP
)
Backend: new unit PM.Handler.Recovery
=====================================
GET /recovery-key/status (auth) -> { configured, created_at? }
POST /recovery-key/setup (auth + CSRF) body {masterPassword, codeHash,
kdfSalt, wrappedKey, wrappedIv}
DELETE /recovery-key (auth + CSRF) -> remove config
POST /recovery-key/redeem (NO auth) body {username, code}
-> session + wrappedKey + wrappedIv + kdfSalt
+ user's current salt + kdfIterations
VerifyMasterPassword() helper handles both legacy 'pbkdf2' and
current 'pbkdf2-sha256' schemes consistently with PM.Handler.Auth.
Setup flow
==========
1. Settings → "Generate recovery code" button (asks master pw via reauth).
2. Client generates: 16-char code + fresh kdf_salt + exports the current
AES key via crypto.subtle.exportKey('raw').
3. Client wraps the raw key under KEK=PBKDF2(code, kdf_salt, 600k)
with a random 12B IV → base64.
4. POSTs to /recovery-key/setup. Server verifies master pw, INSERT-or-
replaces the row (DELETE+INSERT, no UPSERT — same pattern as the
lockout table since FireDAC's UPSERT support is patchy).
5. Confirm modal shows the plaintext code in a monospace, user-select-all
panel. The modal is forcing: "I saved it" button is the only way out.
Modal is the only place the code ever appears — server never sees it.
Redeem flow (forgot master pw)
==============================
1. Auth screen → "Forgot master password? Use a recovery code" link.
2. promptDialog: username, then code (masked input).
3. POST /recovery-key/redeem. Server hashes the typed code, joins with
users by username, ConstantTimeEquals against stored hash. On match:
- deletes the recovery_keys row (single-use)
- issues a fresh session token + CSRF
- returns: { token, csrfToken, salt, kdfIterations, kdfSalt,
wrappedKey, wrappedIv, userId }
4. Client unwraps the AES key with PBKDF2(code, kdfSalt, 600k) → raw bytes
→ importKey('raw') back into a CryptoKey.
5. State is reconstituted from the new session, persistCryptoKey, enterApp.
6. Client immediately opens the Change-master-password modal — the
recovery code is consumed and the account needs a fresh master pw
AND a fresh recovery code (the user generates a new one from Settings).
Backward compat
===============
Recovery is opt-in. Existing users see "No recovery key set" in Settings
until they generate one. No migration needed — the table is created via
CREATE TABLE IF NOT EXISTS at server startup, FK cascade on user delete.
Minor UI additions
==================
- .btn-link CSS class for the auth-screen "Forgot master password?" link.
- Recovery-status label in Settings refreshed on every openSettings()
via GET /recovery-key/status.
Adds the canonical PM feature: let the user pick a new master password
and have every entry transparently re-encrypted under the new key,
without ever exposing plaintext to the server.
Backend endpoint: POST /change-master-password
==============================================
Body:
{
currentMasterPassword, verified against current stored hash
newMasterPassword, basis for the new hash + new client key
newSalt, 64-char hex, client-generated
entries: [{ id, encrypted_password, iv,
totp_secret?, totp_iv? }, ...]
}
Flow:
1. Authenticate + RequireCSRF (caller already logged in).
2. RejectIfAccountLocked — pw change is brute-forceable through a
hijacked session, so it respects the same per-account lockout as
/login.
3. Verify currentMasterPassword against the stored hash. Branches on
hash_algo to handle both legacy 'pbkdf2' and current 'pbkdf2-sha256'.
Wrong pw → RecordFailedAccountAttempt + audit + 401.
4. Compute new auth hash = SHA256(PBKDF2(new_pw, new_salt, 600k)),
always using the current scheme (migration baked in).
5. ATOMIC transaction:
UPDATE users SET password_hash, salt, kdf_iterations, hash_algo
UPDATE vault_entries SET encrypted_password, iv, totp_secret, totp_iv
(per entry)
Any failure → rollback, user stays on the old config.
6. DeleteAllUserSessions — every OTHER session is invalidated so a
leaked old token can't keep working past the rotation. The current
caller's session stays valid.
7. ClearAccountLockout + audit_log entry.
8. Returns { message, salt, kdfIterations }.
Client
======
New modal in index.html (#changeMasterModal) with three password
fields (current / new / confirm) + inline error display. Added a
"Change master password" button in the Settings panel → Account
section. Escape-key handler routes through it like the other modals.
doChangeMasterPassword():
1. Local validation: all fields filled, new ≥ 8 chars, new == confirm,
new ≠ current. Fast failure beats a round trip.
2. randomHexSalt() → 32 secure random bytes, hex-encoded.
3. Derive newKey = PBKDF2(new_pw, new_salt, 600k).
4. Walk state.entries: decrypt password + (optional) TOTP under the
current key, re-encrypt under newKey with fresh random IVs.
One decrypt failure aborts the whole change — better than partial
commit.
5. POST to /change-master-password.
6. On success: swap state.salt + state.cryptoKey, persistCryptoKey,
update sessionStorage, refresh cached ciphertexts in state.entries,
close modal, toast.
7. On 401 / 429 / generic error: show inline error in the modal so
the user can fix and retry without re-typing everything.
Threat model notes
==================
- The current session token stays valid because the new server hash
only invalidates OTHER sessions. Self-logout would be needlessly
disruptive (user already proved knowledge of both pws).
- Server still sees the old + new master pws transiently in /change-
master-password. Same trade-off as /login — eliminating it requires
redesigning to send pre-computed verifiers (SRP-style), tracked
separately.
- The salt rotates with the password — best-practice against any
precomputed dictionary attack tied to the previous salt.
Replaces the plaintext JSON exporter with an encrypted container.
The previous plaintext flow was a known security gap — a backup file
on disk or in a cloud sync folder gave full plaintext access to
every password if accessed by anyone (or anything) other than the
user.
Container format
================
Self-describing JSON:
{
"format": "pm-encrypted-export-v1",
"kdf": "pbkdf2-sha256",
"kdf_iterations": 600000,
"kdf_salt": "<base64 32B>",
"iv": "<base64 12B>",
"ciphertext": "<base64 AES-GCM(payload)>",
"created_at": "<ISO>"
}
payload = same shape as the legacy plaintext exporter (entries array
with site, username, password, folder, tags, favorite, totp_secret,
timestamps), so the round-trip through the JSON importer works
without a separate code path.
Export password
===============
User-chosen, INDEPENDENT of the master password — the export modal
explicitly explains this. Rationale:
- A master-password change doesn't invalidate old backups.
- The backup file can be shared with another person without
revealing the master pw.
- Trade-off: one more password for the user to remember. We assume
they're storing the backup intentionally and can record the pw.
Minimum length 6 enforced client-side.
Flow
====
Export:
1. askReauth(master pw) → server /reauth verifies (defense against
someone reaching the unlocked laptop and dumping the vault).
2. promptDialog(password: true) → export password.
3. Walk state.entries, decrypt each password + TOTP with the vault
key, assemble payload.
4. encryptExportPayload(payload, exportPwd) — random 32B salt,
random 12B IV, PBKDF2 600k, AES-GCM-256.
5. Download the container as
vault-export-YYYY-MM-DD.json.
Import:
1. Read file, detect format. JSON with format === "pm-encrypted-
export-v1" → prompt for the export password.
2. decryptExportContainer → plaintext payload, then JSON.stringify
back into the existing parseEntriesFromJSON path so the rest of
the import flow (preview confirm, bulk encrypt, /entries/bulk-
import) is unchanged.
3. Wrong password → AES-GCM tag fails → "Decryption failed" toast,
user retries.
Other changes
=============
- promptDialog gains a `password: true` option that flips the
confirm input's type so the value is masked on screen.
- Export modal copy in the Settings panel updated to mention the
encrypted format and the independent password.
- The 429-lockout path on /reauth is now handled explicitly in
doExport (was previously falling through to "wrong password").
Backward compatibility
======================
Plaintext JSON exports produced by the previous version still
import — parseEntriesFromJSON doesn't care whether the input came
from a fresh decryption or directly from a plaintext file. The
exporter no longer produces plaintext though; users with old
backups should re-export after upgrading.
Round-trip companion to the existing doExport(). Supports two file
formats with auto-detection (extension + first-char sniff):
JSON
====
Native shape produced by doExport() AND a forgiving fallback for any
flat array of entry objects with site/url + password fields. Accepts:
- { version, exported_at, entries: [...] } (native)
- [{ ... }, { ... }] (flat array)
- mixed keys: site|url|name, username|user|login|email, etc.
CSV
===
RFC-4180-ish parser (~30 lines): quoted fields, escaped "", commas
inside quotes, CRLF line endings. No streaming since password-manager
imports are realistically MB-scale at most.
Heuristic column mapping (case + underscore tolerant) covers the
common exporters out of the box:
Site/URL : name, title, url, site, website, login_uri, login_url
Username : login_username, username, user, login, email
Password : login_password, password, pass, pwd
Folder : folder, group, category, path, collection
Tags : tags, labels (comma/semicolon-split)
Notes : notes, note, comment (short notes joined into tags)
TOTP : login_totp, totp, otpauth, authenticator, two_factor
If the TOTP column holds a full otpauth:// URI it's parsed and only
the secret param is stored — same path used by the slide-over TOTP
field. Invalid base32 TOTP secrets are dropped silently rather than
failing the whole import.
Backend
=======
New endpoint: POST /entries/bulk-import
Body: { entries: [{ site, username, encrypted_password, iv, folder,
tags, totp_secret, totp_iv }, ... ] }
Caps at 10,000 entries per request as a sanity bound. Inserts inside
a single SQLite transaction — partial failure rolls back cleanly, the
user retries from the same source file. Returns { imported: N }.
Rows missing site or ciphertext are skipped within the transaction
(not failed) so one bad row in a 500-entry import doesn't blow up
the whole batch.
Client flow
===========
doImport():
1. Hidden <input type="file" accept=".json,.csv"> picker
2. Read text, detect format, route to parseEntriesFromJSON or CSV
3. confirmDialog preview: count + first 3 sample sites + skipped rows
4. On confirm: encryptImportEntry() each plaintext entry with the
current vault key (reuses encryptPwd / base32Decode validation)
5. Single POST to /entries/bulk-import
6. Reload entries, refresh UI, trigger HIBP scan if enabled
UI
==
Two entry points (mirroring Export):
- Sidebar "Import vault" nav item, next to "Export vault"
- Settings panel "Import" section with descriptive blurb
Both call doImport(). New i-log-in icon added to the SVG sprite (mirror
of i-log-out used by Export).
Limitations
===========
- No de-duplication: importing the same file twice yields duplicate
entries. Trade-off to keep the v1 simple — the user can sort it
out with the existing trash/multi-select UI.
- No password-protected vault formats (Bitwarden encrypted JSON,
KeePass kdbx). Only plaintext exports — same trade-off as
doExport() which produces plaintext JSON.
THE PROBLEM
===========
Before this commit, users.password_hash stored on the server contained
PBKDF2(pw, salt, iters) in hex — the exact same 32 bytes the client
uses as the AES-GCM key to encrypt every entry. Anyone who got hold of
vault.db (filesystem access, backup leak, etc.) had the encryption key
in their hand, no brute force needed. The increased PBKDF2 iteration
count from the previous commit helped against the cipher-text path,
but the easier path was right there in the user row.
THE FIX
=======
Wrap the PBKDF2 output in SHA-256 before storing:
password_hash = SHA256(PBKDF2(pw, salt, iters))
SHA-256 is one-way. The stored hash can still be verified at login
(server recomputes PBKDF2 from the posted master pw, then SHA-256s it,
compares to stored), but the AES key can no longer be recovered from
it. At rest, vault.db only contains an irreversible derivative.
The server still sees pw transiently during /login while computing
the comparison — eliminating that requires a redesigned auth
protocol where the client sends a pre-computed verifier (SRP, OPAQUE,
or simply SHA-256(PBKDF2(pw, salt, iters)) sent from the client).
That's a separate, larger refactor. This commit closes the at-rest
hole, which is the realistic attack surface for vault file leaks.
SCHEMA / MARKER
===============
users.hash_algo distinguishes the two schemes:
'pbkdf2' — LEGACY (raw hex, = AES key)
'pbkdf2-sha256' — CURRENT (SHA-256-wrapped, one-way)
A constant HASH_ALGO_CURRENT replaces the string literal everywhere
to avoid silent drift between the writer and the reader sides.
MIGRATION
=========
Folded into the existing /migrate-kdf endpoint introduced for the
100k→600k iteration bump. Login response now signals migration on
EITHER:
- kdf_iterations < PBKDF2_ITERATIONS_TARGET, OR
- hash_algo != 'pbkdf2-sha256'
The endpoint handles both transitions in one atomic transaction:
UPDATE users SET password_hash = SHA256(PBKDF2(pw, salt, 600k)),
kdf_iterations = 600000,
hash_algo = 'pbkdf2-sha256'
UPDATE vault_entries SET encrypted_password, iv (per entry, if KDF changed)
Idempotency tightened: the "already at target" short-circuit now
requires BOTH conditions, not just the iteration count. Without this,
users who migrated KDF before this commit landed would have been
stuck on the legacy hash format.
CLIENT
======
runKdfMigration() branches on whether the KDF actually changed:
- kdfChange (fromIters !== toIters): re-encrypt all entries with the
new key, send them in the entries array, swap state.cryptoKey on
success. Shows "Vault security upgraded" toast.
- !kdfChange (hash format only): skip the entry re-encryption loop
entirely, send entries: []. Silent — the user didn't perceive a
weakness change worth toasting about.
LOGIN / REAUTH
==============
Both now branch on hash_algo to pick the right verifier:
HASH_ALGO_LEGACY → ConstantTimeEquals(stored, PBKDF2(pw, salt, iters))
HASH_ALGO_CURRENT → ConstantTimeEquals(stored, SHA256(PBKDF2(pw, salt, iters)))
Same constant-time comparison helper as before. Same legacy bcrypt
fallback (still 501-not-implemented).
ALL THREE SCENARIOS AFTER THIS COMMIT
=====================================
1. New register: starts at HASH_ALGO_CURRENT + 600k. No migration ever.
2. Legacy 100k + 'pbkdf2': full migration on next login (hash format
+ iter count + entry re-encryption) in one transaction.
3. Mid-state (already-migrated KDF + still-'pbkdf2'): hash format
upgrade only on next login, no entry re-encryption.
Adds RFC 6238 TOTP (Google Authenticator-style) support to every entry.
The secret is encrypted client-side with the same AES-GCM key as the
password — the server stores opaque ciphertext and never sees the
plaintext base32 secret.
Schema
======
vault_entries.totp_secret TEXT -- AES-GCM ciphertext, base64
vault_entries.totp_iv TEXT -- 12-byte IV, base64
Both NULL when the entry has no 2FA configured. Added via
ApplyMigrations.AddColumnIfMissing so existing vaults migrate cleanly.
Backend
=======
HandleListEntries: includes totp_secret + totp_iv in the response (or
JSON null when not configured).
HandleCreateEntry / HandleUpdateEntry: accept both fields; empty string
in the body → server stores NULL. Clearing the secret removes 2FA
from the entry.
Frontend
========
TOTP primitives (pure crypto.subtle, no external lib):
- base32Decode(s) — RFC 4648, tolerates spaces / lowercase
- generateTOTP(secret) — HMAC-SHA1 + RFC 4226 dynamic truncation
- parseOtpAuthUri(raw) — extracts ?secret from otpauth:// URIs
UI in the slide-over (the canonical entry detail view):
- New "Two-factor (TOTP)" field below the password row.
- Input is password-masked by default with eye-toggle to reveal.
- Pasting a full otpauth:// URI auto-extracts the secret param so the
user can copy directly from a QR-code scanner without manual cleanup.
- X button clears the secret (= removes 2FA on next save).
- Live code panel below: large monospace "123 456" + Copy button
(routes through Bridge.copySecure → secure clipboard + 30s auto-clear).
- Linear progress bar drains over the 30s window, turns red < 5s.
- Refresh tick runs once per second while the slide-over is open;
stops on closeSlideOver to avoid background work.
Entry card meta now shows a "2FA" chip when totp_secret is non-null —
quick visual scan for which accounts have 2FA configured without
opening the slide-over.
Validation
==========
soSave calls base32Decode(secret) before encrypting to refuse obviously
broken input. Otherwise garbled base32 would save fine and only fail
in the code panel next time.
Migration interaction (KDF 100k→600k)
=====================================
KNOWN MINOR ISSUE: /migrate-kdf only re-encrypts encrypted_password+iv,
not totp_secret+totp_iv. In practice this is harmless because:
1) KDF migration runs immediately after login on legacy accounts —
before the user has a chance to add a TOTP secret.
2) New accounts start at 600k iterations, no migration ever needed.
A legacy user who somehow added a TOTP between login and the
background migration completing would end up with a TOTP encrypted
under the old key. The fix (extend /migrate-kdf to re-encrypt TOTP
fields too) is a one-line follow-up if anyone hits the edge case.
HIBP integration
================
Opt-in (default OFF) password breach check via the Have I Been Pwned
range API. The full master / entry password never leaves the machine —
only the first 5 characters of its SHA-1 hash. HIBP returns ~500
candidate suffixes; the client matches its own suffix locally.
UI:
- New "Check passwords against breach database (HIBP)" toggle in
Settings → Security with an explainer hint about k-anonymity.
- On enable: background batch scan of all entries, results cached in
state.hibpResults keyed by entry id. Concurrency capped at 6 to
avoid hammering HIBP / hitting browser connection limits.
- Entry cards show a red "Pwned" chip + breach count in the tooltip
when count > 0. New i-alert icon added to the SVG sprite.
- Auto-scan triggered after every enterApp() when the toggle is on.
Functions added to app.js:
- sha1Hex(text) — crypto.subtle wrapper
- hibpCheckPassword(plaintext) — single-password check, returns count
- hibpCheckAllEntries() — batched scan over state.entries
The "Add-Padding: true" header is sent on every range request to defeat
the response-size side-channel (HIBP adds 800-1000 random extra entries
so an observer counting bytes can't narrow the prefix queried).
CSP tightening
==============
Audited the served HTML: zero <script> tags inline, only the external
js/app.js. Removed 'unsafe-inline' from script-src — real XSS defense.
Kept 'unsafe-inline' on style-src for now because index.html contains
inline style="" attributes and app.js calls element.style.cssText
extensively. Refactoring to CSS classes is a separate cleanup. Style
injection alone cannot execute code, so the residual risk is bounded
to visual manipulation in a single-user loopback app.
Added api.pwnedpasswords.com to connect-src as the only allowed
external origin (required by the HIBP feature above). Default still
'self' — everything else stays loopback.
Before:
script-src 'self' 'unsafe-inline';
style-src 'self' 'unsafe-inline';
connect-src 'self';
After:
script-src 'self';
style-src 'self' 'unsafe-inline';
connect-src 'self' https://api.pwnedpasswords.com;
Bumps the PBKDF2-SHA256 iteration count from 100,000 (OWASP 2017) to
600,000 (OWASP 2023). 6x slowdown on every brute-force attempt against
either the server-stored auth hash OR the AES-GCM ciphertext of the
entries — both currently use the same PBKDF2 output (see KNOWN ISSUE
below for why that's another problem to fix later).
Schema
======
users.kdf_iterations INTEGER DEFAULT 100000
Per-user iteration count. Legacy rows predating the column default
to 100k via the DEFAULT clause. New accounts insert 600k explicitly.
Migration flow
==============
Atomic from the user's perspective. No partial state ever persisted.
1. /login (or /reauth):
server reads users.kdf_iterations and verifies the master pw at
that count. Login succeeds at the legacy strength. Response now
includes kdfIterations (current) and optionally kdfMigration =
{ target: 600000 } when an upgrade is recommended.
2. Client:
derives the AES key at the OLD count to decrypt current entries
(state.cryptoKey). enterApp() loads the vault normally.
3. runKdfMigration() (background, after enterApp):
- derives the NEW key at target iterations
- decrypts every entry with the old key
- re-encrypts every entry with the new key + fresh random IVs
- POSTs { masterPassword, entries: [...] } to /migrate-kdf
4. /migrate-kdf (new endpoint):
- verifies the master pw against the OLD hash
- in a single transaction:
UPDATE users SET password_hash = pbkdf2(pw, salt, 600k),
kdf_iterations = 600000
UPDATE vault_entries SET encrypted_password, iv (per entry)
- on any failure: ROLLBACK. User stays at legacy config, retries
at next login. No half-migrated state possible.
5. Client (post-commit):
swaps state.cryptoKey to the new key, persists it, updates the
cached ciphertext in state.entries, shows a "Vault security
upgraded" toast.
Idempotency: server's /migrate-kdf short-circuits with "Already at
target" if users.kdf_iterations >= PBKDF2_ITERATIONS_TARGET.
Race conditions: two concurrent migrations from two tabs both
recompute the SAME new key (deterministic PBKDF2). The losing
transaction's entries get re-encrypted with the winning one's IVs,
but both clients can decrypt because the keys are identical.
KNOWN ISSUE (not fixed by this commit)
======================================
The server's password_hash IS the client's AES key, in hex form —
both sides compute PBKDF2(pw, salt, iters) and store/use the same
32 bytes. This means a stolen vault.db gives the attacker the
encryption key directly, without needing to brute-force anything.
The 100k → 600k bump still helps because the AES-GCM ciphertext
itself is also a brute-force target, but the architectural fix
(server stores SHA256(aes_key) instead of aes_key in hex) is a
separate concern that needs its own migration.
Other changes
=============
- HandleRegister: new accounts insert kdf_iterations=600000.
- HandleReauth: response upgraded to JSON with kdfIterations
+ optional kdfMigration. Unlock path now also triggers migration.
- SendAuthSuccess: extended signature, all callers updated.
- deriveKey(pwd, saltHex, iterations) in app.js: iterations param
required, defaults to 100000 for back-compat with any legacy caller.
Existing protection was per-IP only (login_attempts table). On a loopback
deployment everyone hits 127.0.0.1, so the per-IP counter is mostly
ornamental — the real attacker is on the same machine. Adds a second
defense layer that tracks failures per username with an exponential
backoff schedule.
Schema:
account_lockouts (username TEXT PK, failed_count INT,
locked_until DATETIME, last_attempt_at, last_attempt_ip)
Backoff after threshold (4+ failures):
1, 2, 3 failures → no lockout (grace window for typos)
4th → 60 s
5th → 5 min
6th → 15 min
7th → 1 h
8th → 6 h
9th and beyond → 24 h (capped)
Counter resets to 0 on successful login or reauth. Old non-locked rows
older than 30 days are pruned by CleanupExpired alongside the existing
sessions / audit_log / login_attempts cleanups.
Wiring:
- HandleLogin / HandleReauth both check RejectIfAccountLocked() before
touching the users table. Lockout responses are 429 with JSON body
{ error, retry_after } and a Retry-After header.
- Failed attempts are recorded against the username even when the user
doesn't exist, preventing account enumeration via differential
"is this account locked?" probes.
- PBKDF2 hash comparison was already constant-time (ConstantTimeEquals);
no change there.
Client (js/app.js):
- api() now preserves response status + body on Error so callers can
distinguish 429-lockout from other errors.
- New showLockoutCountdown(seconds) renders a live "Account locked —
try again in Xm Ys" message in #authHint, disables #loginBtn until
the countdown reaches 0, then re-enables it.
- doLogin / doUnlock both branch on err.status === 429 + retry_after
to call showLockoutCountdown instead of a generic error toast.
Known limitation: an attacker can DoS-lock arbitrary usernames by
spamming /login with that name. This is intentional — the alternative
(per-(username,IP) tracking) would let attackers enumerate accounts.
DoS-lock is acceptable; auth bypass is not.
Introduces the Delphi 12 FMX backend (PMServer) that hosts the embedded
WebView2 vault on 127.0.0.1, and a native bridge between JS and Delphi
that wires three privacy-focused features:
1. Secure clipboard
Copying a password registers the Win32 "ExcludeClipboardContentFromMonitorProcessing"
format alongside CF_UNICODETEXT, so Win+V clipboard history never sees
the value. Auto-clears after 30s via TTimer. Bridge.copySecure() in
app.js routes all password/username/secret copy paths through the
native layer when running inside the Delphi WebView2 (falls back to
navigator.clipboard for the PHP standalone).
2. Tray icon (X-to-tray when server running)
Closing the dev panel hides both the form HWND and the TFMAppClass
per-process proxy window that owns the FMX taskbar entry — the form's
HWND alone is not the taskbar-visible one in FMX (took some iteration
to discover). Tray menu: Open, Lock vault, Quit. Clipboard is force-
cleared on minimize as extra safety. First-time minimize fires a
balloon notification so the user knows the app is still running.
3. Auto-lock on Windows session lock (Win+L)
wtsapi32.dll!WTSRegisterSessionNotification on a dedicated message-only
window. On WM_WTSSESSION_CHANGE / WTS_SESSION_LOCK, the bridge calls
ExecuteJavaScript('lockVault()'). Same path used by the tray "Lock vault"
menu item.
Bridge architecture:
- JS → Delphi via cmd:// URLs intercepted in OnBeforeNavigate
(pattern lifted from DeskInsight Monaco). Currently exposes
cmd://clipboard/copy?text=...&clear=... and cmd://clipboard/clear.
- Delphi → JS via TTMSFNCWebBrowser.ExecuteJavaScript with guarded
calls (typeof check) so the bridge degrades cleanly if app.js isn't
loaded yet.
Files:
- Source/PM.Bridge.pas (new) — TSecureClipboard + TPMBridge
- UMainForm.pas/.fmx — bridge wiring, FormCloseQuery intercept, tray
callbacks (BridgeTrayRestore / BridgeLockRequest / BridgeQuit)
- js/app.js — Bridge object, 5 navigator.clipboard sites migrated to
Bridge.copySecure with PHP-compatible fallback, Bridge.onTrayRestore
handler that resets the auto-lock timer
.gitignore extended with Delphi build artifacts (*.dcu, Win32/, Win64/,
__history/, __recovery/, *.identcache, *.dsk, *.local, etc.) so source
checkouts stay clean.
Dark theme: neutral dark grays (#0e1015 bg, #6b7280 accent). Light theme: white cards, better contrast. Cards now have subtle shadows for separation. All hardcoded rgba(59,130,246) replaced with rgba(var(--accent-rgb), ...). Arrow key handler skips when add/edit modal is open. Detail view now uses getFilteredEntries() instead of getFilteredEntries(true) to respect folder selection.
Previous approach always started from the lowest-indexed selected entry, causing the range to snap to anchor on arrow reversal. Now arrowFocus tracks the moving end; arrowAnchor stays fixed. Non-Shift arrow resets both to the new position.
Shift+arrow now uses arrowAnchor (set on first arrow press) instead of lastSelectedId (which was updated on each non-Shift move). Fixes 'only 2 entries' bug. Detail view: fixed inverted auth check for ArrowLeft/Right; general arrow handler now excludes detail view.
Use single-byte rejection sampling: generate byte, reject if >= largest multiple of charset length, then modulo. Eliminates bias from c.charAt(arr[i] % c.length).
Ctrl+N and Ctrl+T cannot be intercepted by Chrome (browser-level shortcuts). Replaced with Alt+N and Alt+T which work reliably. Updated keyboard shortcuts help.
Switched from e.code (physical key position) to e.key (character value) which is more reliable across keyboard layouts and browsers. Also removed duplicate Ctrl guard line.
document-level keydown bubbling phase is too late for browser-level shortcuts Ctrl+N (new window) and Ctrl+T (new tab). Moved handlers to a window.addEventListener('keydown', ..., true) capture-phase handler that fires before the browser chrome acts. Removed redundant document-level copies.
Moved Ctrl+N and Ctrl+T handlers before the generic Ctrl guard with their own e.preventDefault() and early return, matching the pattern used for Ctrl+A. Removed them from the old else-if chain and combined preventDefault block.
Ctrl+A was selecting page text instead of entries. Switched from e.code === 'KeyA' (layout-dependent) to e.key === 'a'. Moved handler before the generic Ctrl+key else-if chain and before the combined preventDefault block, with its own e.preventDefault() and early return. Handles both 'a' and 'A' key values.
Ctrl+A auth check was inverted (!hidden instead of hidden), preventing selection in vault. Added Delete key shortcut: moves selected entries to trash (or batch permanent delete in trash view). Added Del to keyboard shortcuts help.
Document mousedown handler was clearing selectedIds when clicking 'Yes' (button element). Since mousedown fires before click, selectedIds was empty by the time the async callback read it, causing no-op deletes but still showing the toast. Added .batch-confirm-overlay to the mousedown exclusion list so clicking the confirm dialog won't trigger clearSelection().
Replace showCenterConfirm with showBatchConfirm(btn,message,callback) that creates a transparent modal overlay (z-index 9999) to block background clicks, positions the confirm dialog beside the trigger button. Updated all callers (batchDelete, batchPermanentDelete, permanentDelete, emptyTrash) to pass the clicked button. Added CSS for batch-confirm-overlay backdrop.
Add showCenterConfirm() function for dialogs without button anchor. Replaced all standard confirm() calls in batchDelete, batchPermanentDelete, permanentDelete, emptyTrash. Batch permanent delete shows single custom confirm with entry count, no nested dialogs. permanentDelete() accepts silent param to suppress UI for batch operations. Order cleanup added to permanentDelete.
Trash drop handler now uses noToast + single message. Restore button icon changed to ♻️. Removed stale .view-toggle reference in click-outside handler. Added safety check in reorder drop handler to prevent entries being lost from order array.
Batch delete/restore now shows single toast with count instead of one per entry. delEntry() and restoreEntry() accept noToast param. Added Ctrl+A to select all visible entries. Updated keyboard shortcuts help.
Replace view toggle buttons with dropdown menu containing all 7 views. Card view: 2-column grid with bigger cards. Grouped view: entries grouped by folder with sticky headers. Detail view: single entry at a time, large text, prev/next navigation. Arrow key navigation for detail view.
Added user-select:none and -webkit-user-drag:none to #entriesContainer.
Also added a dragstart listener that prevents default for any element
that is not [draggable='true'], stopping the browser from creating drag
ghosts of selected table text/headers during rectangle selection.
The entriesContainer has no padding so cards fill the grid edge-to-edge;
there is no clickable empty space at the left/top/bottom of the card area
within entriesContainer. Moved rect-selection initiation into the
document-level mousedown handler so it fires when clicking on ANY
non-interactive element inside .vault (padding, grid gaps, etc.).
Interactive elements (buttons, inputs, folders bar, toolbar, etc.)
still clear selection as expected.
Document mousedown handler clears selection when clicking on any element
that is not an entry card, the batch bar, or an open modal/settings menu.
Works in tandem with entriesContainer rect-select handler: when clicking
inside entriesContainer on empty space the container handler runs first
(setting rectState.active=true) so the document handler skips;
when clicking outside entriesContainer the document handler fires alone
and calls clearSelection().
Zaki