Bot Feedback Loop

The Bot Feedback Loop is a way for customers to send Cloudflare direct feedback in the case of Bot Management potentially scoring a request incorrectly. When a customer submits a False Negative or a False Positive report, Cloudflare manually analyzes this data and uses it as a training dataset for our next Machine Learning model.

​​ Availability

Bot Feedback Loop is available for Enterprise Bot Management customers. Visit Plans for more information.

​​ False Positive

A false positive can happen if Cloudflare scores a request from a person using a browser, mobile application or desktop application in the automated or likely automated range.

​​ False Negative

If Cloudflare is unable to detect a portion of automated traffic on your site, submitting a False Negative report will help us catch it in the future.

​​ Subtypes

​​ Submit a report

1. Log in to the Cloudflare dashboard Open external link and select your account and domain.
2. Go to Security > Bots.
3. Apply one or more bot score filters.
4. Select Report incorrect data and fill out the form.
5. Select Submit.

​​ Via the API

​​ Create a feedback report

curl 'https://api.cloudflare.com/client/v4/zones/{zone_id}/bot_management/feedback' \
--header "X-Auth-Email: <EMAIL>" \
--header "X-Auth-Key: <API_KEY>" \
--header "Content-Type: application/json" \
--data '{
  "type": "false_positive",
  "description": "Legitimate customers having low score",
  "expression": "(cf.bot_management.score le 46 and ip.geoip.asnum eq 132892 and http.host eq \"api-discovery.theburritobot.com\" and cf.bot_management.ja3_hash eq \"3fed133de60c35724739b913924b6c24\")",
  "first_request_seen_at": "2022-08-01T00:00:00Z",
  "last_request_seen_at": "2022-08-10T00:00:00Z",
  "requests": 100,
  "requests_by_score": {
    "1": 50,
    "10": 50
  },
  "requests_by_score_src": {
    "heuristics": 25,
    "machine_learning": 75
  },
  "requests_by_attribute": {
    "topIPs": [
      {
        "metric": "10.75.34.1",
        "requests": 100
      }
    ],
    "topUserAgents": [
      {
        "metric": "curl/7.68.0",
        "requests": 100
      }
    ]
  }
}'

# Output
Null

​​ List feedback reports

curl 'https://api.cloudflare.com/client/v4/zones/{zone_id}/bot_management/feedback' \
--header "X-Auth-Email: <EMAIL>" \
--header "X-Auth-Key: <API_KEY>"

// Output
[
  {
    "created_at": "2022-08-19T00:05:24.749712Z",
    "type": "false_positive",
    "description": "Legitimate customers having low score",
    "expression": "(cf.bot_management.score le 46 and ip.geoip.asnum eq 132892 and http.host eq \"api-discovery.theburritobot.com\" and cf.bot_management.ja3_hash eq \"3fed133de60c35724739b913924b6c24\")",
    "first_request_seen_at": "2022-08-01T00:00:00Z",
    "last_request_seen_at": "2022-08-10T00:00:00Z",
    "requests": 100,
    "requests_by_score": {
      "1": 50,
      "10": 50
    },
    "requests_by_score_src": {
      "heuristics": 25,
      "machine_learning": 75
    },
    "requests_by_attribute": {
      "topIPs": [
        {
          "metric": "10.75.34.1",
          "requests": 100
        }
      ],
      "topUserAgents": [
        {
          "metric": "curl/7.68.0",
          "requests": 100
        }
      ]
    }
  }
]

​​ API Fields

requests_by_score

{
  "1": 50,
  "10": 50
}

requests_by_score_src

{
  "machine_learning": 75,
  "heuristics": 25
}

requests_by_attribute

  {
    "topIPs": [
      {
        "metric": "10.75.34.1"
        "requests": 100
      }
    ],
    "topUserAgents": [
      {
        "metric": "curl/7.68.0",
        "requests": 100
      }
    ]
  }

​​ Expression fields

​​ Recommendations when submitting a report

When you submit a report, use the filters available in the Bot Analytics dashboard to ensure that your report includes only the traffic that received an incorrect score. In addition to filtering by a score (required), you may want to filter by user-agent, IP, ASN or JA3 to more precisely highlight the section of traffic that was scored incorrectly.

If you are not certain if some traffic received an incorrect score, keep this traffic in the report.

We appreciate any comments you wish to leave in the description field that might help our team better understand these requests in the context of typical traffic to your domain.

​​ Recommendations after submitting a false positive

After submitting a false positive, you can explicitly allow the traffic if you are confident that this traffic source cannot be used for abuse in the future. To allow traffic, you can create a WAF custom rule with a Skip the remaining custom rules action that matches the characteristics of your false positive report. We recommend any skip rule that you create uses the most narrow possible scope, including restricting the request methods and URIs that the expected traffic has access to, to limit potential abuse.

• Allowing a JA3/JA4 fingerprint: If you want to allow access to a stable software client that does not come from a dedicated IP, you can do so by looking up the JA3 fingerprint(s) used by that client in the Bot Analytics dashboard, and creating a WAF custom rule to allow traffic based on that JA3 fingerprint. JA3 fingerprints will only match a client’s TLS library, so be cautious in looking for both overlap with other clients and with variation based on the operating system.
  
  Cloudflare does not recommend relying on JA3 rules for mobile applications that may be abused. If you have questions about how to securely allow traffic from your mobile application, please contact your account team.

• Allowing an IP address: Only use an IP address to allow traffic if the IP is a dedicated resource that belongs only to the traffic source you wish to allow.
  If the traffic you want to allow shares an IP with other traffic sources, or if the IP changes frequently, consider an alternative to allowing by IP address.

​​ Recommendations after submitting a false negative

After submitting a false negative report, you can explicitly block or rate-limit the incorrectly scored traffic using a combination of characteristics such as IP address, JA3 fingerprint, ASN, and user-agent. Before blocking or rate-limiting based on JA3 fingerprint, please use Bot Analytics to confirm that fingerprint is not being used by legitimate traffic sources.