log-leak
Fail your test run when personal data or secrets show up in your logs.
This line passes code review and every test, then sends your users' email addresses to Datadog for the next two years:
log.info(f"created user {user.email}")
log-leak is a pytest plugin. While your tests run, it reads every log record your code writes and checks it for emails, phone numbers, NIC numbers, card numbers, passwords, tokens and API keys. Your tests already go through the code paths that log, so they find the leaks for you:
$ pytest
........................................ [100%]
=================================== log-leak ===================================
app/users.py:6 LL009 email An email address was logged.
in create_user(), logger 'app.users' at INFO
message: created user <email>
first seen in: tests/test_users.py::test_create
also in 3 more tests
fix: Log an internal user ID instead.
app/auth.py:31 LL006 credential A value under a key like password, secret or token was logged.
in login(), logger 'app.auth' at WARNING
message: login failed for {'username': 'amal', 'password': '<credential>'}
template: login failed for %s
first seen in: tests/test_auth.py::test_wrong_password
fix: Log only the fields you need, or mask sensitive keys before logging.
log-leak: found 2 leaks at 2 call sites, logged by 5 tests.
40 passed in 1.31s
The report shows where the log call is, never the leaked value itself, so the CI log
doesn't become a second leak. Fake test data counts as a leak: if a test logs
amal@example.com, the same line logs real addresses in production.
Install
pip install pytest-log-leak # or: uv add --dev pytest-log-leak
That's it. pytest loads the plugin on its own, so the next pytest run is checked.
What it finds
| ID | Name | Finds |
|---|---|---|
| LL001 | private-key | PEM private key blocks |
| LL002 | jwt | JSON Web Tokens |
| LL003 | api-key | AWS, GitHub, Stripe, Slack, OpenAI, Anthropic, Google and SendGrid keys |
| LL004 | url-credentials | Passwords inside URLs, e.g. postgresql://app:secret@db/prod |
| LL005 | auth-header | Bearer and Basic Authorization header values |
| LL006 | credential | Values under keys like password, secret, token, api_key |
| LL007 | card-number | Card numbers (checked against card prefixes and the Luhn checksum) |
| LL008 | nic | Sri Lankan NIC numbers, old and new format |
| LL009 | Email addresses | |
| LL010 | phone | Sri Lankan mobile numbers, and any number written as +<country code>... |
It checks the log message, the messages of any exception logged with
log.exception(...), and fields passed with extra={...}.
When a leak is fine
Some logged values are meant to be public, like a support address. Silence one call with a comment on its first line, or alone on the line above:
log.info("contact %s", SUPPORT_EMAIL) # log-leak: ignore[LL009]
# log-leak: ignore -- every detector, for this call only
log.info(
"sent receipt to %s",
order.email,
)
pytest --log-leak-explain LL009 prints why a detector exists and how to fix what it
finds. --log-leak-explain all lists them.
Adopting it in an existing project
A big project may have dozens of leaks on day one. Record them in a baseline file and fix them over time:
pytest --log-leak-update-baseline # writes .log-leak-baseline.json; commit it
pytest # now only new leaks fail the run
Baseline entries are keyed by file, function, logger and detector, not line number, so
editing code above a log call doesn't make an old leak look new. When an entry stops
showing up, log-leak tells you, and the next --log-leak-update-baseline after a full run
drops it.
Configure
In pytest.ini, or [tool.pytest.ini_options] in pyproject.toml:
[pytest]
# Lowest level checked. Production usually logs INFO.
log_leak_level = INFO
# false: report leaks without failing the run.
log_leak_fail = true
# Detectors to skip everywhere.
log_leak_ignore = LL010
# Loggers to skip, with their children.
log_leak_ignore_loggers = sqlalchemy.engine
# Where the baseline lives.
log_leak_baseline = .log-leak-baseline.json
Turn it off for one run with pytest -p no:log_leak.
GitHub Actions
Nothing to set up. When GITHUB_ACTIONS=true, log-leak also prints each leak as an
annotation, so it shows up on the pull request diff at the logging line.
How accurate is it
A checker that fails your build has to be almost never wrong, or people turn it off. The detectors are tuned against Loghub: 32,000 real log lines from 16 systems, including Hadoop, Spark, OpenSSH, Android, macOS and Windows.
The first run gave 79 hits. 67 were false positives: HDFS block IDs that pass the card
checksum, times followed by process IDs that looked like phone numbers, compact timestamps
that looked like NIC numbers, and object descriptions like token=Token{a64f992}. Each is
fixed and kept as a regression test. The other 12 hits are real email addresses: two
personal ones in the macOS logs (one inside an iCloud URL, logged 10 times) and Red Hat's
build address in a Linux kernel banner.
Run it yourself with uv run python bench/loghub.py. Loghub contains almost no secrets, so
it measures false positives, not how many real leaks are caught. That side is covered by the
test suite.
Limits
- pytest-xdist is not supported yet. With
-n, log-leak prints a warning and checks nothing. - It sees records that reach the root logger. A logger with
propagate = Falseis not checked. - It finds what your tests exercise. Code paths with no test are not checked.
caplog.set_level(...)in a test raises the level for that test, so lower-level records in it aren't created and can't be checked.
License
MIT
Metadata
Release files for pytest-log-leak 0.1.0
For a detailed explanation of source distributions (sdists) and built distributions (wheels), please see the package formats documentation.
Source distribution (sdist)
| File | Size | Uploaded | Provenance | Actions |
|---|---|---|---|---|
| pytest_log_leak-0.1.0.tar.gz | 29.5 kB | Without provenance | Details for pytest_log_leak-0.1.0.tar.gz |
Built distribution (wheel)
| File | Interpreter | ABI | Platform | Actions Reset |
|---|---|---|---|---|
| pytest_log_leak-0.1.0-py3-none-any.whl | Python 3 | none | any | Details for pytest_log_leak-0.1.0-py3-none-any.whl |
Total release size: 46.6 kB
Release files / pytest_log_leak-0.1.0.tar.gz
| Download URL | pytest_log_leak-0.1.0.tar.gz |
|---|---|
| Size | 29.5 kB |
| Tags | Source |
|
SHA-256 checksum How to use checksums |
919044ff16ac934256d0507efa01d2f2f027dd50e46c8f107ad70645b94af3e3
|
|
BLAKE2b-256 checksum How to use checksums |
4ed1e81dd305c2855a8f1a240aeb767dc6bf5fc4e7da624924da7a6eb2ed7601
|
| Upload date | |
|
Uploaded using Trusted Publishing? What is trusted publishing? |
Yes |
| Uploaded via |
uv/0.13.6 {"installer":{"name":"uv","version":"0.13.6","subcommand":["publish"]},"python":null,"implementation":{"name":null,"version":null},"distro":{"name":"Ubuntu","version":"24.04","id":"noble","libc":null},"system":{"name":null,"release":null},"cpu":null,"openssl_version":null,"setuptools_version":null,"rustc_version":null,"ci":true}
|
Release files / pytest_log_leak-0.1.0-py3-none-any.whl
| Download URL | pytest_log_leak-0.1.0-py3-none-any.whl |
|---|---|
| Size | 17.1 kB |
| Tags | Python 3 |
|
SHA-256 checksum How to use checksums |
32ed8bc7efca5bb7aafda14e088fbfc625b6f0013d4e8d8dcfeaeebb8267fe1e
|
|
BLAKE2b-256 checksum How to use checksums |
57aab0c7d0789bfeb09d07276a2a9d57d07bc0da1d9e02e2686af15cbe742bd1
|
| Upload date | |
|
Uploaded using Trusted Publishing? What is trusted publishing? |
Yes |
| Uploaded via |
uv/0.13.6 {"installer":{"name":"uv","version":"0.13.6","subcommand":["publish"]},"python":null,"implementation":{"name":null,"version":null},"distro":{"name":"Ubuntu","version":"24.04","id":"noble","libc":null},"system":{"name":null,"release":null},"cpu":null,"openssl_version":null,"setuptools_version":null,"rustc_version":null,"ci":true}
|
Security for pytest-log-leak 0.1.0
Provenance
Provenance describes where a file came from. On PyPI, provenance is shared via attestations, which provide a verifiable record of the build or publishing details. View details, limitations and caveats.
No provenance found
None of the files in this release were uploaded with provenance metadata.
A note on package security
PyPI is a repository for distributing packages, not a service for auditing them. While third-party security researchers and our staff do their best to detect and remove malicious packages, PyPI cannot guarantee the quality, security or suitability of any given package.
To learn more about protecting yourself from malicious actors, visit our package security documentation.