password-breach-exposure
Use after any major data-breach headline or when auditing which of your accounts are compromised — checks passwords against the Have I Been Pwned breach corpus using k-anonymity (only a 5-character SHA-1 prefix ever leaves the machine), maps each hit to what the breach actually leaked (passwords vs contact info vs payment), and produces a severity-ranked rotation queue with credential-stuffing blast-radius math instead of a vague "change everything."
Password Breach Exposure
A breach drops. The advice is always "change your passwords" — all of them? in what order? which ones actually matter? Most people either panic-rotate one password or do nothing. Neither survives contact with an attacker running credential stuffing, where a LinkedIn-2012 password meets a Gmail login box within hours.
This skill makes the exposure concrete and the response ordered:
- Check — test specific passwords against the real breach corpus via Have I Been Pwned's Pwned Passwords API using k-anonymity: the script hashes locally (SHA-1), sends only the first 5 hex characters, and matches the suffix locally. The password itself never leaves the machine — not even in hashed-full form.
- Assess — map each hit to what that breach class actually leaked (plaintext passwords vs bcrypt vs email-only vs payment data). A password in a bcrypt breach is urgent; the same password in a contact-scrape is cosmetic.
- Prioritize — rotation queue ranked by (leaked credential
type × reuse × account value), with a blast-radius estimate:
exposed uses = reuse count × stuffing hit-rate.
When to Use
- A breach is in the news and you had an account there
- Periodic hygiene audit ("which of my ~10 base passwords are burned?") — quarterly is reasonable
- After finding an old password written down / in a screenshot / in an email to yourself
- Building or rebuilding a password-manager vault: test base passwords BEFORE reusing them anywhere
- Helping a family member who "uses the same password everywhere" and wants to know how bad that actually is
Don't use for: choosing a NEW password's strength (different problem — length/passphrase math), phishing/URL forensics (different skill), or enterprise SIROM work. This is personal triage.
The Privacy Architecture (why this is safe)
The Pwned Passwords API is designed for exactly this flow:
SHA-1(password) = 21BD1... (computed locally)
prefix sent to API = "21BD1" (5 hex chars — that's ALL)
API returns = every suffix in its corpus starting 21BD1
(typically 500-800 hash suffixes + counts)
local match = compare your suffix against the list
- The API learns nothing usable: a 5-char prefix covers ~1M+ possible hashes; your password is one of hundreds in the response
- No account, no email, no identifiers are sent — password check and email/breach check are SEPARATE queries (the script never combines them in one request)
- Offline fallback: if the network is unavailable, the script falls back to a bundled catalog of ~40 major breaches with leaked-field classes and dates, so assessment mode still runs
The Rotation Queue Logic
Each exposed password gets a priority score:
severity = leaked_type_weight (plaintext 100 / bcrypt 65 /
unsalted-hash 80 / contact-only 10)
priority = severity × (1 + reuse_count) × account_weight
(email/bank/phone = 3, shopping = 1.5, forums = 1)
Output is a numbered queue: rotate #1 first, with the reason attached ("plaintext leak + 4 known reuses + guards an email account"). Anything contact-only goes to the bottom as "low-value rotation."
Pitfalls
- Testing passwords by pasting them into random "breach checker" websites. Unknown operators = you just handed over the credential. Only k-anonymity against a known endpoint, or fully local checks.
- Treating every breach as equal. A 2012 leak of bcrypt hashes with no salts cracked by 2024 is still dangerous, but an email-only marketing leak is not a password event. The leaked-field class drives urgency.
- Rotating the exposed password everywhere EXCEPT the old forum it leaked from. The stale copy re-burns the new one if you reuse. The queue includes the source account.
- Email-based panic. "Your email appeared in a breach" emails describe the email appearing — often just contact data. Check WHAT leaked before spending a weekend on rotation.
- Forgetting the username side. If the email itself leaked, credential stuffing also tries email+top-1000 passwords. A strong unique password survives that; 2FA on email/bank makes it moot — the report flags missing 2FA on high-value targets when you list them.
- One-and-done. Old dumps keep circulating. Re-run the check quarterly and after every major headline.
Verification Checklist
- Passwords tested only via k-anonymity (verify: request URL contains a 5-char prefix, nothing longer)
- Each hit mapped to leaked-field class + breach date
- Reuse count honestly estimated per password
- Rotation queue generated with priorities, #1 actionable now
- High-value accounts flagged for 2FA, not just rotation
- Breach sources dated — pre-2016 plaintext leaks are still fully live threats
Quick Recipe
# interactive, password hidden on input
python3 scripts/breach_exposure.py check
# audit a set of base passwords non-interactively (careful: shells
# keep history — prefer the interactive mode)
python3 scripts/breach_exposure.py check --pw "old-base-1" --pw "old-base-2"
# map exposure to your accounts and build the rotation queue
python3 scripts/breach_exposure.py assess --email you@example.com \
--inventory my_accounts.txt
# offline catalog of major breaches + what they leaked
python3 scripts/breach_exposure.py catalog
my_accounts.txt: one account password-nickname weight per
line (e.g. gmail.com base-A 3, old-forum base-A 1) — the
script cross-references nicknames against checked passwords and
emits the ranked queue.
- Downloads
- Security audit
- Review
- Last updated
- 4w ago
- Current version
- v1.0.0
- License
- MIT-0
