Prefilter policies
A prefilter policy is a security policy that
-
applies to incoming network connections before other security policies,
-
evaluates traffic using layer 3 and layer 4 criteria such as protocol, source IP address, destination IP address, and port, and
-
enables early decision-making to reduce processing requirements and enhance device performance.
Prefilter policy actions
Use prefilter policies to implement these main actions:
-
Improve performance—The sooner you exclude traffic that does not require inspection, the better. Specifically, you can perform these actions.
-
Offload trusted flows—If your device supports flow offload, use the prefilter Fastpath action to identify flows you trust to be eligible for offloading to NIC processing. No further processing, including inspection, is needed for these connections. Refer to Best practices for fastpath prefiltering and Large flow offloads.
-
Block unwanted connections—For example, if you want to block entire subnets, protocols, or TCP/UDP ports, configure this in the prefilter policy.
-
-
Rezone plain text encapsulated tunnels so you can handle them as a unit in the access control policy. Some protocols, such as GRE, create plain-text tunnels that include many connections. By rezoning a tunnel, you assign it a tag, which you can select in access control rules as a source or destination zone. This allows you to apply customized intrusion or other rules to the tunneled traffic and provides granular insight and control.
Note
VPN or other encrypted tunnels are not plain-text tunnels and cannot be rezoned.
You must assign each access control policy a prefilter policy. However, Cloud-Delivered Firewall Management Center includes a pre-defined prefilter policy that you can use.
Default prefilter policy
A default prefilter policy only analyzes plain-text encapsulated tunnels. It is automatically assigned when you create a new access control policy. It cannot be deleted from the system.
Default prefilter policy characteristics
Every access control policy must have an assigned prefilter policy. For convenience, the system includes a default prefilter policy that is automatically assigned when you create a new access control policy.
You can instead change the default policy to block all plain-text tunnels. See Configure the default action.
If you want to do any other customization, including adding prefilter or tunnel rules, you must create your own prefilter policy and assign it to the appropriate access control policies, as explained in Assign access control policy to devices.
Prefilter policy rule order
Prefilter policy rule order is a traffic processing mechanism that determines which tunnel rules and prefilter rules handle network traffic first and applies the default action for unmatched tunnels.
Rule processing sequence
In a prefilter policy, tunnel rules, prefilter rules, and a default action handle network traffic:
-
Tunnel and prefilter rules—First, rules in a prefilter policy handle traffic in the order you specify. Tunnel rules match specific tunnels only. Prefilter rules have a wider range of constraints.
Rules are matched in top-down order, first match wins. Thus, rule order is critical. For example, you might want to block an entire subnet except for two specific IP addresses on the subnet. In that case:
-
The first rule should specify the two allowable IP addresses as the source network, and the rule action would be Analyze. Connections from these two IP addresses would match this rule, and the next rule would not be processed.
-
The next rule should specify the entire subnet as the source network, with Block for the rule action.
For more information, see Tunnel vs prefilter rules.
-
-
Default action (tunnels only)—If a tunnel does not match any rules, the default action handles it. The default action can block these tunnels or continue access control on their individual encapsulated connections (Analyze action).
There is no default action for nonencapsulated traffic. If a nonencapsulated connection does not match any prefilter rules, the system continues with access control. These connections might subsequently be blocked by other policies, or ultimately allowed.
To change the rule order, you can:
-
Drag and drop the rule to the required position.
-
Right-click the rule and choose Cut, then find the correct position, right-click, and choose Paste Above or Paste Below.
-
Edit the rule and use the Insert control to change its position.
Tunnel vs prefilter rules
Choose to configure a tunnel rule or a prefilter rule based on the type of traffic you want to match and the actions or analysis you intend to perform.
|
Characteristic |
Tunnel Rules |
Prefilter Rules |
|---|---|---|
|
Primary function |
Quickly fastpath, block, or rezone plaintext, passthrough tunnels. |
Quickly fastpath or block any other connection that benefits from early handling. |
|
Encapsulation and port/protocol criteria |
Encapsulation conditions match only plaintext tunnels over selected protocols: GRE, IP-in-IP, IPv6-in-IP, Teredo. |
Port conditions can use a wider range of port and protocol constraints than tunnel rules; see Port, protocol, and ICMP code rule conditions (concept). |
|
Network criteria |
Tunnel endpoint conditions constrain the endpoints of the tunnels you want to handle. |
Network conditions constrain the source and destination hosts in each connection. |
|
Direction |
Bidirectional or unidirectional (configurable). Tunnel rules are bidirectional by default, so they can handle all traffic between tunnel endpoints. |
Unidirectional only (nonconfigurable). Prefilter rules match source-to-destination traffic only. Return traffic for allowed connections is also permitted. |
|
Rezone sessions for further analysis |
Supported, using tunnel zones; see Using tunnel zones to apply access control at the tunnel level. |
Not applicable. |
How the system processes plain-text tunnels
Some protocols, such as GRE, create plain-text (not encrypted) tunnels that contain (or encapsulate) multiple inner connections. These connections might flow between discontinuous networks. These tunnels are especially useful for routing custom protocols over IP networks, IPv6 traffic over IPv4 networks, and so on.
Summary
A plain-text tunnel has two sets of IP headers:
-
Outer headers: Specify tunnel endpoint addresses for prefilter policy handling
-
Inner headers: Contain actual connection endpoint addresses for granular inspection
Workflow
The system processes plain-text tunnels through these stages:
-
The system receives a plain-text tunnel with dual IP header structure.
- Outer header—An outer encapsulation header specifies the source and destination IP addresses of the tunnel endpoints—the routed interfaces of the network devices on either side of the tunnel.
- Inner headers—Inner payload headers specify the source and destination IP addresses of the encapsulated connections' actual endpoints. There are separate headers for each connection contained within the tunnel.
- The prefilter policy processes the outer tunnel headers to determine handling approach. The prefilter policy uses outer headers to handle traffic. You can block or fast path entire plaintext, passthrough tunnels at this stage. If you do not want to allow plain-text tunnels at all, you can set the default action in the prefilter policy to block all plain-text tunnels. This block does not apply to encrypted tunnels such as VPN.
- The system exposes inner headers for tunnels that pass prefilter analysis. Any tunnels that you analyze (that is, not block or fast path) have these inner headers exposed for use by the rest of the security policies and other features such as QoS. These features use the innermost detectable level of headers to ensure the most granular level of inspection and handling possible. This ensures that bad actors cannot hide illegitimate traffic within a plain-text tunnel. Alternatively, you can use tunnel zones to re-tag tunneled traffic so that you can write access control rules specifically for select tunnels. This lets you customize intrusion inspection or other layer 7 filtering to connections contained with a tunnel. See Using tunnel zones to apply access control at the tunnel level.
Using tunnel zones to apply access control at the tunnel level
A tunnel zone represents certain types of plaintext, passthrough tunnels that you explicitly tag for special analysis. Although a tunnel zone is not an interface object, you can use it as an interface constraint in some configurations.
Tunnel zone implementation
As explained in How the system processes plain-text tunnels, once a plain-text tunneled connection passes through the prefilter policy, the access control policy applies its rules to each connection contained within the tunnel.
However, you might want to apply the same policy to all the connections within a tunnel. For example, you might have a special intrusion policy you want to use for tunneled connections.
To make it possible to apply access control rules to all connections within a tunnel, you must use the prefilter policy to re-zone the tunnel. Rezoning applies a tunnel zone tag to the plain-text tunnel. You can then use that tunnel zone instead of a security zone when you define the source interface criterion of an access control rule.
Because a tunnel zone does not contain interfaces, the source interface through which the tunnel enters the device is not relevant. You can apply a tunnel zone tag to multiple tunnels that enter the device through different interfaces and access control treats them all in the same way.
Note |
Connections in rezoned tunnels do not match security zone constraints in an access control rule. After you rezone a tunnel, access control rules match encapsulated connections with their newly assigned tunnel zone, not the original security zone. |
See Rezone tunnels for customized inspection for a walkthrough of a tunnel zone implementation, and a discussion of the implications of rezoning without explicitly handling rezoned traffic.
Large flow offloads
A large flow offload is a hardware-based traffic processing feature that
-
handles certain fastpathed traffic in the NIC hardware rather than in the Firewall Threat Defense software
-
results in higher throughput and lower latency for data-intensive applications such as large file transfers, and
-
includes both static flow offload (based on prefilter policy configuration) and dynamic flow offload (based on trust and other criteria).
Supported models and offload types
For the following models, certain traffic that you configure to be fastpathed by a prefilter policy is handled by the hardware (specifically, in the NIC), not by the Firewall Threat Defense software:
-
Firepower 4100/9300
-
Secure Firewall 3100 (static only)
-
Secure Firewall 4200 (static only)
-
Secure Firewall 6100
Offloaded flows continue to receive limited stateful inspection, such as basic TCP flag and option checking. The system can selectively escalate packets to the firewall system for further processing if necessary.
Two types of flow offload are available:
-
Static flow offload: Offloads flows based on prefilter policy configuration that applies the Fastpath action.
-
Dynamic flow offload: By default, devices offload flows based on other criteria, including trust.
Applications that can benefit from offloading large flows include:
-
High Performance Computing (HPC) Research sites, where the firewall is deployed between storage and high compute stations. When one research site backs up using FTP file transfer or file sync over NFS, the large amount of data traffic affects all connections. Offloading FTP file transfer and file sync over NFS reduces the impact on other traffic.
-
High Frequency Trading (HFT), where the firewall is deployed between workstations and the Exchange, mainly for compliance purposes. Security is usually not a concern, but latency is a major concern.
Flows that can be offloaded include:
-
(Static flow offload only.) Connections that are fastpathed by the prefilter policy.
-
Standard or 802.1Q tagged Ethernet frames only.
-
Cluster asymmetric flows (Secure Firewall 4200). For asymmetric flows, cluster redirect lets the forwarding node offload flows to hardware. When traffic for an existing flow is sent to a different node, then that traffic is redirected to the owner node over the cluster control link. Because asymmetric flows can create a lot of traffic on the cluster control link, letting the forwarder offload these flows can improve performance.
-
(Dynamic flow offload only):
-
Inspected flows that the inspection engine decides no longer need inspection. These flows include:
-
Flows handled by access control rules that apply the Trust action and are based on security zone, source and destination network and port matching only.
-
TLS/SSL flows that are not selected for decryption using a decryption policy.
-
Flows that are trusted by the Intelligent Application Bypass (IAB) policy either explicitly or due to exceeding flow bypass thresholds.
-
Flows that match file or intrusion policies that result in trusting the flow.
-
Any allowed flow that no longer needs to be inspected.
-
-
IPS preprocessor inspected flows:
-
SSH and SMTP.
-
FTP preprocessor secondary connections.
-
Session Initiation Protocol (SIP) preprocessor secondary connections.
-
-
Intrusion rules that use keywords (also referred to as options)
-
Important |
For details, exceptions, and limitations to the above, see Restrictions for flow offload implementation. |
Static flow offload configuration
To offload eligible traffic to hardware, create a prefilter policy rule that applies the Fastpath action. Use prefilter rules for TCP/UDP, and tunnel rules for GRE.
(Not recommended.) To disable flow offload, use FlexConfig:
-
Disable static (and thereby dynamic) flow offload: no flow-offload enable
-
Re-enable flow offload: flow-offload enable
-
Disable cluster redirection: no flow-offload cluster-redirect
For more information, see Cisco Secure Firewall ASA Series Command Reference.
Dynamic flow offload configuration
Dynamic flow offload is enabled by default. However, dynamic offload occurs only if static flow offload is enabled, regardless of whether prefiltering is configured.
To disable dynamic flow offload, use the device CLI:
-
Disable dynamic offload: configure flow-offload dynamic whitelist disable
-
Re-enable dynamic offload: configure flow-offload dynamic whitelist enable
Restrictions for flow offload implementation
Device limitations
The feature is supported on these devices:
-
Secure Firewall 3100 (static only)
-
Secure Firewall 4200 (static only)
-
Secure Firewall 6100
-
Firepower 4100/9300
Flows that cannot be offloaded
Do not attempt to offload these flow types:
-
Any flows that do not use IPv4 addressing, such as IPv6 addressing.
-
Flows for any protocol other than TCP, UDP, and GRE.
Note
PPTP GRE connections cannot be offloaded.
-
Flows on interfaces configured in passive, inline, or inline tap mode. Routed and switched interfaces are the only types supported.
-
(3100/4200/6100.) Offload based on inner header for tunnelled flows.
-
(3100/4200.) Multi-instance offload.
-
Flows that require inspection by Snort or other inspection engines. In some cases, such as FTP, the secondary data channel can be offloaded although the control channel cannot be offloaded.
-
IPsec and TLS/DTLS VPN connections that terminate on the device.
Note
The Secure Firewall 3100 and Secure Firewall 4200 platforms are exceptions to this restriction. These platforms support flow-offload for IPsec and DTLS VPN connections terminating to the device. For detailed information on this feature and its implementation, please refer to the release notes specific to the Secure Firewall 3100 and Secure Firewall 4200 platforms.
-
Flows that require encryption or decryption. For example, connections decrypted due to a decryption policy.
-
Multicast flows in routed mode. They are supported in transparent mode if there are only two member interfaces in a bridge group.
-
TCP Intercept flows.
-
TCP state bypass flows. You cannot configure flow offload and TCP state bypass on the same traffic.
-
Flows tagged with security groups.
-
Reverse flows that are forwarded from a different cluster node, in the case of asymmetric flows in a cluster.
-
Centralized flows in a cluster, if the flow owner is not the control unit.
-
Flows that include IP options cannot be dynamically offloaded.
Additional limitations
Consider these operational restrictions when implementing flow offload:
-
Flow offload and Dead Connection Detection (DCD) are not compatible. Do not configure DCD on connections that can be offloaded.
-
If more than one flow that matches flow offload conditions are queued to be offloaded at the same time to the same location on the hardware, only the first flow is offloaded. The other flows are processed normally. This is called a collision. Use the show flow-offload flow command in the CLI to display statistics for this situation.
-
Dynamic flow offload disables all TCP normalizer checks.
-
Although offloaded flows pass through FXOS interfaces, statistics for these flows do not appear on the logical device interface. Thus, logical device interface counters and packet rates do not reflect offloaded flows.
Cluster redirection
When traffic for an existing flow is sent to a different node, then that traffic is redirected to the owner node over the cluster control link. Because asymmetric flows can create a lot of traffic on the cluster control link, offloading these flows can improve performance, for example, for distribited site-to-site VPN mode. This feature is supported for these environments:
-
Secure Firewall 4200
-
Routed and transparent mode
-
Spanned EtherChannel mode
Conditions for reversing offload
After a flow is offloaded, packets within the flow are returned to the firewall operating system for further processing if they meet these conditions:
-
They include TCP options other than Timestamp.
-
They are fragmented.
-
They are subject to Equal-Cost Multi-Path (ECMP) routing, and ingress packets move from one interface to another.
-
A change to the routing table applies to the flow. A packet is returned to the device to determine the new route.
Prefilter vs access control policy
Prefilter and access control policies both allow you to block and trust traffic. In a prefilter policy, the "trust" functionality is called "fastpathing" because it skips more inspection. If you do not configure a custom prefilter policy, you can only approximate, but not replicate, prefilter functionality using Block and Trust rules placed early in the access control policy.
|
Characteristic |
Prefilter Policy |
Access Control Policy |
|---|---|---|
|
Primary function |
Quickly fastpath or block certain types of plaintext or passthrough tunnels, or tailor subsequent inspection to encapsulated traffic. Fastpath or block any other connections that benefit from early handling. |
Inspect and control all network traffic, using simple or complex criteria, including contextual information and deep inspection results. |
|
Sequence within access control |
First The system matches traffic to prefilter criteria before all other access control configurations. Thus, fast-pathing in a prefilter policy skips many other policies and inspections. |
Last Other policies that might inspect a connection, including Security Intelligence and decryption, are applied before access control rules are evaluated. |
| Rule actions |
Fewer You can stop further inspection (Fastpath and Block) or allow further analysis with the rest of access control (Analyze). |
More Access control rules have a larger variety of actions, including monitoring, deep inspection, block with reset, and interactive blocking. See Access control rule actions. |
|
Bypass capability |
Fastpath rule action Fastpathing traffic in the prefilter stage bypasses all further inspection and handling. This includes:
|
Trust rule action Traffic trusted by access control rules is only exempt from intrusion inspection and discovery. If another policy that requires inspection is applied to a connection that is eventually trusted, the connection is inspected anyway. Trust means that intrusion and file policies are not applied to the connection. |
|
Rule criteria |
Limited Rules in the prefilter policy use simple network criteria: IP address, VLAN tag, port, and protocol. For tunnels, tunnel endpoint conditions specify the IP address of the routed interfaces of the network devices on either side of the tunnel. |
Robust Access control rules use network criteria, but also user, application, requested URL, and other contextual information available in packet payloads. Network conditions specify the IP address of source and destination hosts. |
|
IP headers used (tunnel handling) |
Outermost Using outer headers allows you to handle entire plaintext, passthrough tunnels. For nonencapsulated traffic, prefiltering still uses "outer" headers—which in this case are the only headers. |
Innermost possible For a nonencrypted tunnel, access control acts on its individual encapsulated connections, not the tunnel as a whole. |
|
Rezone encapsulated connections for further analysis |
Rezones tunneled traffic Tunnel zones allow you to tailor subsequent inspection to prefiltered, encapsulated traffic. |
Uses tunnel zones Access control uses the tunnel zones you assign during prefiltering. See Using tunnel zones to apply access control at the tunnel level. |
|
Connection logging |
Fastpathed and blocked traffic only. Allowed connections may still be logged by other configurations. |
Any connection |



Feedback