Applications and services

Configure NetScaler to delete or pass the upgrade header

A new parameter passProtocolUpgrade is added to the HTTP profile to prevent attacks on the back-end servers. Depending on the state of this parameter, the upgrade header is passed in the request sent to the back-end or deleted before sending the request.

• If the passProtocolUpgrade parameter is enabled, then the upgrade header is passed to the back end. The server accepts the upgrade request and notifies it in its response.
• If this parameter is disabled, then the upgrade header is deleted and the remaining request is sent to the back end.

The passProtocolUpgrade parameter is added to the following profiles.

• nshttp_default_profile - enabled by default
• nshttp_default_strict_validation - disabled by default
• nshttp_default_internal_apps - disabled by default
• nshttp_default_http_quic_profile - enabled by default

We recommend that you set the passProtocolUpgrade parameter to disabled.

Set the passProtocolUpgrade parameter by using the CLI:

At the command prompt, type the following:

set ns httpProfile <name> [-passProtocolUpgrade ( ENABLED | DISABLED )]

Set the passProtocolUpgrade parameter by using the GUI:

1. Navigate to System -> Profiles -> HTTP Profiles.
2. Create or edit an HTTP profile.
3. Select the Pass Protocol Upgrade checkbox.

Configure NetScaler to drop invalid HTTP requests

We recommend that NetScaler is configured with strict checking and enforcement of HTTP requests to prevent invalid HTTP requests passing through virtual servers. To do so, bind an in-built HTTP profile, nshttp_default_strict_validation, to one or more virtual servers using the following command on the CLI:

show ns httpProfile (Shows the available http profile (default+user configured profiles))

set lb vserver <vserver name> -httpProfileName nshttp_default_strict_validation
<!--NeedCopy-->

We recommend that customers using this option test the changes in a staging environment before releasing it to production.

Protection against the HTTP desync attacks is enabled by default on the strict HTTP validation profile (nshttp_default_strict_validation). Use the strict profile for all the client-facing entities.

For more information on HTTP request smuggling attacks and its mitigation, see the support article NetScaler - HTTP Request Smuggling Reference Guide.

Configure protection against HTTP Denial of Service attacks

The NetScaler firmware supports limited countermeasures against HTTP Denial of Service attacks, including ‘slow-read’ type attacks. You can configure these features by using the nsapimgr utility from the NetScaler shell prompt:

• small_window_threshold (default=1)
• small_window_idle_timeout (default=7 sec)
• small_window_cleanthresh (default=100)
• small_window_protection (default=Enabled)

The default settings are adequate for preventing the HTTP Denial of Service attacks, including slow-read attacks, however, some tuning of the parameters might be required for other attacks.

To protect against such attacks, adjust the small_window_threshold property upward by using the following nsapimgr command from the NetScaler shell prompt:

$ nsapimgr –ys small_window_threshold=<desired value>

Note:

The small_window_threshold desired value can be set based on the incoming traffic pattern in the deployment. The acceptable range is from 0 to 2^32.

You can verify the protection against HTTP Denial of Service attacks by monitoring the following counters with nsconmsg –d stats command from the NetScaler shell prompt:

• nstcp_cur_zero_win_pcbs: This counter tracks the number of PCBs that currently have a low window value.
• nstcp_err_conndrop_at_pass: This counter is incremented when NetScaler detects that, while passing packets through from one side to the other, it has exceeded the nscfg_small_window_idletimeout value.
• nstcp_err_conndrop_at_retx: This counter is incremented when the time that lapses during retransmission exceeds the nscfg_small_window_idletimeout value.
• nstcp_cur_pcbs_probed_withKA: This counter tracks the number of PCBs in the surge queue that are probed with a KA probe.

We recommend that customers using this option test the changes in a staging environment before releasing it to production.

Configure NetScaler to defend against TCP spoofing attacks

The following commands can be used to help protect back-end servers against TCP spoofing attacks:

set ns tcpProfile profile1 -rstWindowAttenuate ENABLED -spoofSynDrop ENABLED

Done

set lb vserver lbvserver1 -tcpProfileName profile1

Done
<!--NeedCopy-->

We recommend that customers using this option test the changes in a staging environment before releasing it to production.

Configure RFC 5961 compliance

Starting with NetScaler release 14.1-4.x, the -rfc5961Compliance option is available in the TCP profile. This option includes spoofing protection and covers both -rstWindowAttenuate and -spoofSynDrop. Enable -rfc5961Compliance to get all these protections without setting each option separately.

For more information, see Sample TCP Configurations.

Configure NetScaler to accept specific HTTP headers

It is possible to configure the NetScaler to accept only specific HTTP headers. This can be accomplished by adding a rewrite action to restrict network traffic with specific, defined HTTP headers from being passed to the back-end server.

The following global rewrite action sends only network traffic with headers such as Host, Accept, and test to the server:

add rewrite action act1 replace_all q/HTTP.REQ.FULL_HEADER.after_str("\r\n")/     q{TARGET.REGEX_SELECT(re/(iu)^(Host|Accept|test):.*\r\n/) ALT ""} -pattern q{re/(U).+:.+r\n/}

add rewrite policy pol1 HTTP.REQ.IS_VALID act1

bind rewrite global pol1 100
<!--NeedCopy-->

Configure header size for HTTP headers

We recommend you to maintain a value suitable for your business needs for maxHeaderLen and maxHeaderFieldLen HTTP header size parameters. The default value is set to 24820, which ensures compatibility with applications that require large HTTP headers.

set httpprofile http-prof -maxHeaderLen <header-length> -maxHeaderFieldLen <header-field-length>
<!--NeedCopy-->

The maxHeaderLen parameter controls the total header size and maxHeaderFieldLen controls the size of individual header fields. These parameters are available to enhance security by limiting the attack surface for header-based exploits, such as buffer overflow or header injection attacks. Set the header size on the HTTP profile that is configured on the virtual server and the service group or the service that is bound to the virtual server.

While it is possible to configure these parameters to a lower value, doing so might result in legitimate requests being dropped if the application uses headers exceeding the configured limits. Therefore, the administrators must carefully assess their application requirements before lowering these values.

Configure close-notify

A close-notify is a secure message that indicates the end of SSL data transmission. In compliance with RFC 5246: The client and the server must share knowledge that the connection is ending to avoid the truncation attack. Either party can initiate the exchange of closing messages by sending a close_notify alert. Any data received after a closure alert is ignored, unless some other fatal alert has been transmitted, each party is required to send a close_notify alert before closing the write side of the connection. To ensure that audit events are captured for TLS termination events log on to the CLI as a superuser or sysadmin and run the following commands:

set ssl parameter -sendCloseNotify y

save ns config
<!--NeedCopy-->

Secure Management GUI, NITRO API, and RPC communication

To secure the management GUI, NITRO API, and RPC communication on NetScaler and NetScaler Gateway, you must set -maxclient in NetScaler for internal HTTP/HTTPS service.

set service nshttpd-gui-127.0.0.1-80 -maxclient $max_client{}

set service nshttpd-vpn-127.0.0.1-81 -maxclient $max_client{}

set service nshttps-127.0.0.1-443 -maxclient $max_client{}

<!--NeedCopy-->

For more information on setting the -maxclient value and the NetScaler versions that support this setting, see Configure the internal HTTP GUI service by using NetScaler GUI, or NetScaler CLI, or NetScaler NITRO APIs.