Skip to content

Latest commit

 

History

History
62 lines (41 loc) · 1.98 KB

cli-slowloris.md

File metadata and controls

62 lines (41 loc) · 1.98 KB
Raw

Is my server Slowloris-proofed?

Slowloris attack attempts to break an HTTP server by sending partial HTTP request, which contains never finish HTTP header lines:

GET / HTTP 1.1  # this is the only line required to start an HTTP request
HOST example.com
User-Agent my-user-agent
Header-Name Header-Value
# ... keep sending gibberish header name & value lines

Since the HTTP request is never ended, vulnerable server keeps the connection open. As a result, server side resources like memory, file descriptor will be consumed.

Invulnerable server should close the connection after a specified time if the request is unable to read completely.

We can use turtle to validate if an HTTP endpoint is Slowloris-proofed:

  • if the server is immune to the attack, we should see closed events, and the total number of requests should be more than the test connections;
  • if the server is vulnerable to the attack, the connections will be kept opened until test finished.

Validating via CLI

NOTE We can start a test server with turtle-proof. For setup guide, please see turtle proof server.

  1. Start a vulnerable server:
$ turtle-proof
2023/09/04 11:33:04 INFO server started turtle-proof.addr=127.0.0.1:8889
  1. Launch the test with the slowloris sub-command:
$ turtle slowloris http://127.0.0.1:8889 --http-send-gibberish

We should see output similar to below, where the number of connections stays at 100 without closing.

  1. Start a invulnerable server:
$ turtle-proof --scenario=proof
2023/09/04 11:36:49 INFO server started turtle-proof.addr=127.0.0.1:8889
  1. Launch the test again
$ turtle slowloris http://127.0.0.1:8889 --http-send-gibberish

Now, we should see many closing / reopening events like this: