Redacting data

To maintain data privacy, Next-Gen WAF redacts sensitive data from requests before they reach the platform backend.

Selective data transfer and redaction

The Next-Gen WAF agent filters requests locally to determine if they contain an attack. Only requests that are marked as attacks or anomalies are then sent to the platform backend after additional filtering and sanitizing are done. Once the agent identifies a potential attack or anomaly in a request, the agent sends only the individual parameter of the request which contains the attack payload, as well as a few other non-sensitive or benign portions of the request (e.g., client IP, user agent, or URI). The entire request is never sent to the platform backend. Additionally, specific portions of the request are automatically redacted and never sent to the backend, including tokens, credentials, and known patterns such as credit card and social security numbers.

JSON API payloads

Next-Gen WAF automatically parses JSON key-value pairs and treats them like request parameters. The following sample requests demonstrate how redactions work within the context of a request.

The initial request:

POST /request HTTP/1.1
Content-Length: 72
Content-Type: application/json
Host: api.example.com
{"user":"user@api.example.com","password":"<script>alert(1)</script>mypassword","zip":94089}

What's sent to the Next-Gen WAF:

POST /request HTTP/1.1
Host: api.example.com

password=

The initial request:

POST /request HTTP/1.1
Content-Length: 72
Content-Type: application/json
Host: api.example.com

{"user":"user@api.example.com","password":"mypassword","zip":"<script>alert(1)</script>94089"}

What's sent to the Next-Gen WAF:

POST /request HTTP/1.1
Host: api.example.com

zip=<script>alert(1)</script>

Sensitive headers

Next-Gen WAF redacts the following from requests:

• Explicit names: authorization, x-auth-token, cookie, set-cookie
• Any names that contain: -token, -auth, -key, -sess, -pass, -secret
• Query strings from referer and location

The initial request:

POST /example?sort=ascending HTTP/1.1
Host: example.com
User-Agent: Mozilla/5.0 (Macintosh; Intel Mac OS X 10.10; rv:35.0)
Accept: text/html, application/xhtml+xml
Content-Length: 57
Cookie: foo=bar

sensitive=hunter2&foobar=<script>alert(1)</script>&page=3

What’s sent to the Next-Gen WAF:

POST /example HTTP/1.1
Host: example.com
User-Agent: Mozilla/5.0 (Macintosh; Intel Mac OS X 10.10; rv:35.0)

foobar=<script>alert(1)</script>

Sensitive parameters

If a request contains an attack or anomaly, and also contains sensitive data in commonly-used parameter names, Next-Gen WAF will redact the entire contents of the sensitive parameter. These parameters include:

• api_key
• password
• passwd
• pass
• pw
• user
• login
• loginid
• username
• email
• key
• id
• sid
• token
• request_token
• access_token
• csrfmiddlewaretoken
• oauth_verifier
• confirm_password
• password_confirmation

The initial request:

POST /example HTTP/1.1

username=<script>alert("jsmith")</script>

What’s sent to the Next-Gen WAF:

POST /example HTTP/1.1

username=[redacted]

The control panel clearly displays which parameters have been redacted. Redacted parameters are replaced with the word REDACTED highlighted in yellow.

Sensitive patterns

Next-Gen WAF automatically redacts known patterns of sensitive information, which includes the following:

• Credit card numbers: values like 4111-1111-1111-1111 become 0000-0000-0000-0000
• Social security numbers: values like 078-05-1120 become 000-00-0000
• GUIDs: values like 3F2504E0-4F89-41D3-9A0C-0305E82C3301 become 0000000-0000-0000-0000-000000000000
• Bank account (IBAN) numbers: values like DE75512108001245126199 become AA00aaaa0000000

The initial request:

POST /example HTTP/1.1

credit_card_example=<script>alert("4111-1111-1111-1111")</script>

What’s sent to the Next-Gen WAF:

POST /example HTTP/1.1

credit_card_example=<script>alert("0000-0000-0000-0000")</script>

Within the control panel we clearly display which patterns have been redacted. Redacted patterns are replaced with the word REDACTED highlighted in yellow.

Custom redactions

In addition to the redactions listed above, you can also specify additional fields to redact from requests. For example, if your password field is named foobar instead of password, that field can be specified for redaction.

Creating custom redactions

When you have a sensitive field that is not filtered out by default, you can create a custom field redaction:

Editing custom redactions

To edit a custom redaction, complete the following steps:

Deleting custom redactions

Transparency

To allow for easy verification of what the agent sends to the backend, we provide a way to view all agent to backend communication.

Verifying in the control panel

To verify our agents are correctly filtering and sanitizing requests, we provide a raw log of data that’s sent from our agents:

Verifying with the agent

You can also verify directly from the agent itself by setting the debug-log-uploads agent configuration option. For example, if you want to log all agent uploads in formatted JSON, add the following line to your agent configuration file (by default at /etc/sigsci/agent.conf):

debug-log-uploads = 2