What actually happens to a password here
Exactly what's implemented, described plainly rather than reassuringly — including the parts that are limitations, not just the parts that sound good. If something here changes in the code, this page should change with it.
What gets generated
Two different generators are behind the one form, and they don't use the same quality of randomness.
Word-basedThe word comes from the database's own randomization; the digit and symbol come from PHP's general-purpose rand(). Fine for everyday onboarding passwords — not a cryptographic random source.Random charsEvery character, and the order they're shuffled into, comes from PHP's random_int() — a cryptographically secure generator, throughout.Every generated password is checked against Have I Been Pwned's breach database using k-anonymity: only the first 5 characters of its SHA-1 hash ever leave the server. The password itself — or even its full hash — is never transmitted anywhere.
Encryption at rest
Creating a Secure Link encrypts the text with AES-256-CBC before it's written to the database, with a fresh random IV generated per link — the same text encrypted twice never produces the same ciphertext. The link itself is a 256-bit random token, not a guessable ID.
One key, held by the server, shared by every link. The encryption key is a single value read from an environment variable — not stored in a web-reachable file — but it's the same key for every link ever created, not a unique key per link. That protects the data if, say, only a database backup leaks. It does not protect against someone who compromises the application server itself, since they'd have the key too.
This payload also isn't separately signed or authenticated (no HMAC) — unlike this site's own .intunewin packaging, which does add one. In practice this only matters to someone who already has write access to the database, at which point simpler attacks already exist.
Access control
Anyone can generate passwords and create encrypted links. Both are rate-limited per IP address to stop the tool being scripted; logged-in users and approved IPs are exempt from the limit.
Sending anything — by SMS or by email — is restricted to the IP allowlist (Admin → IP Allowlist), and being logged in does not get around it. Visitors from other IPs don't see the sending controls at all, and the backend independently re-checks the allowlist on every send, so calling the endpoint directly rather than through the page gains nothing.
Those same actions also require a valid CSRF token, generated fresh per session and checked on every request. Before this, several of them were plain GET requests with real side effects (including one that put the password itself in the URL) — a malicious page loaded by someone on an approved network could have silently triggered an SMS send or a link creation with no interaction beyond the page loading. That's closed now: everything state-changing requires a same-session token and a real POST.
Delivery methods
Send as Plain Text SMS puts the actual password directly in the message body, sent through the org's SMS gateway over HTTPS. It's the simplest option, but the password then exists as plain text on the recipient's phone, in their SMS history, and passed through a third-party provider — ordinary SMS was never designed as a secure channel.
Send Link via SMS/Email is different by design: only the URL is ever sent. The password stays encrypted in the database until someone actually opens that link, and the link can be capped to as few as one view and expire in as little as an hour.
Every send is logged with only a masked phone number or email (last 4 digits, or first letter + domain) and a success/failure flag — never the message content itself.
Retention — how long is anything kept
A daily scheduled task deletes a Secure Link's row once it's no longer servable — expired, or its view limit used up — typically within 24 hours of that point (cron_purge_expired_links.php, confirmed running as a cron job). Before that point, the encrypted row exists for exactly as long as the expiry setting chosen when the link was created (1 hour to 7 days).
The recipient doesn't have to wait on either of those. Once a link is opened, a Delete Now button sits right under the revealed content — clicking it removes that row immediately, on the spot, regardless of how many views or how much time was left on it. There's no undo and no confirmation email; the token simply stops resolving to anything from that point on.
The activity log (who sent what, when, to which masked number) is not covered by that cleanup and is still kept indefinitely — it holds no passwords, only masked metadata, but there's no automatic retention limit on it yet.
The plaintext password itself is never written to disk anywhere by this app — not in the activity log, not in any file. The only places it ever exists are: briefly in server memory while being generated or encrypted, inside the encrypted database row (until the daily purge removes it), and wherever the delivery channel itself — the SMS gateway, the recipient's phone, an email inbox — keeps it once sent. All of those are outside this tool's control.
Quick answers
This page describes the code as it actually works today, not a policy statement — if the encryption, access control, or retention behavior changes, update this page in the same change. An accurate "here's what it doesn't do yet" is more useful than a vague "it's secure."
Back to the Password Generator