> ## Documentation Index
> Fetch the complete documentation index at: https://docs.dagworks.io/llms.txt
> Use this file to discover all available pages before exploring further.

# Security & Data

> Information on security and data collected.

## Security

We take security very seriously and we use industry standard security practices to ensure your data is safe.

If you see an issue or have questions, email us at [security@dagworks.io](mailto:security@dagworks.io).

### Advised practices

We recommend you follow the following practices to ensure your data is secure:

1. Secure your API Keys by storing them in an appropriate secrets management solution.
2. API Keys are bound to individual users, and allow writing to projects that a user has been granted access to.
   Each user should have their own API key(s). A user can have many API Keys. *Currently there is no programmatic API read
   access to your data.*
3. If an API Key is exposed, the organization admin, and/or individual user can revoke
   the key (and create a new one) via [clicking the key icon](https://www.app.dagworks.io/dashboard/settings) on the
   left hand navigation bar. Email us at [security@dagworks.io](mailto:security@dagworks.io) if you need help.

### Sensitive Data Scrubbing

If you sent data to DAGWorks that you didn't want to, email [security@dagworks.io](mailto:security@dagworks.io),
providing the project ID(s) and the run ID(s) to delete.

## Data Collected

### DAGWorks CLI

We currently do not capture usage data from the DAGWorks CLI.

### DAGWorksTracker

Via the DAGWorks Tracking Adapter (DAGWorksTracker), we capture the following information:

1. The Hamilton code defining the DAG. This is required.
2. DAG Execution telemetry. This is required.
3. Summary statistics of any tabular/vector data observed, in addition to any python primitive return values. This is required.
4. Coming soon - python environment dependencies used by a DAG run. This is required.
5. Telemetry on DAGWorksTracker usage. Optional. You can opt-out of telemetry by setting the environment variable `DW_DISABLE_TRACKING=true`.

We will add the ability to add custom filtering/processing in the future, but for now we only store the above.

If you have any questions about what we do and don't store, please contact us at [support@dagworks.io](mailto:support@dagworks.io).

### Function return types we introspect

We introspect the following function return types automatically:

| Type             | What we capture                                                        |
| ---------------- | ---------------------------------------------------------------------- |
| Pandas DataFrame | Summary statistics                                                     |
| Pandas Series    | Summary statistics                                                     |
| Python primitive | The actual value for int, float, string, booleans.                     |
| Polars DataFrame | Summary statistics                                                     |
| Polars Series    | Summary statistics                                                     |
| Numpy Arrays     | Summary statistics                                                     |
| Python dicts     | Actual values or summary statistics based on values in the dictionary. |
| Python list      | Actual values or summary statistics based on values in the list.       |

### Masking sensitive data

To mask sensitive data it shouldn't be serializable. One way is to wrap it in an object. For example,
In the case of `api_keys` you can create a class like the following:

```python theme={null}
class Secret:
    def __init__(self, value):
        self._value = value

    def __repr__(self):
        return "********"

    @property
    def value(self):
        return self._value
```

And then use it like this:

```python theme={null}
def my_function(api_key: Secret) -> pd.DataFrame:
    client = Client(api_key=api_key.value)
    ...  # do stuff
    return result
```