ASL firewall

From Atomicorp Wiki
Revision as of 17:56, 22 March 2013 by Mshinn (Talk | contribs)

Jump to: navigation, search

Contents

Introduction

ASL has two different ways you can configure the firewall on your system:

1) Fast/Easy Mode

This configuration system should be the one most new users are familiar with. This system lets you quickly define a list of ports to allow or deny in a list. For most uses the Fast/Easy mode is what you will want to use.

2) Advanced Firewall Rule Manager

This is a full featured netfilter GUI. This will allow you to define arbitrary chains, actions, modules and other advanced iptables features. This is recommended for advanced users.

How the firewall works

The ASL stateful packet inspection firewall works much like other firewalls. It inserts rules into the Linux kernel which govern how the kernel will handle network traffic. All firewalls work in a "first match wins" manner. That means rules are checked in order, and the first rule to apply "wins". For example, if you have two firewall rules in this order:

  1. Allow traffic on port 80
  2. Block all traffic

Lets say you have a connection comes in on port 80. The firewall will run through the rules in the order you configure. In the example above, the first rule says to allow traffic on port 80. This connection is to port 80, so the rule will match and the traffic will be allowed and processing will stop. Theres nothing else for the firewall to do, its allowed the traffic so its done.

However, lets say in the example above if a packet comes in on port 23. The firewall will run through the rules again and this time rule 1 will not match. This is a request to access port 23, not port 80. So processing will continue. Moving on to rule 2, this rule says to block all traffic. So, this rule will match, because all traffic means all traffic, including this request to port 23. Therefore this rule will match, the firewall will block the request, and processing will stop because theres nothing else to do, the connection was blocked.

Heres another important example. Lets say your firewall rules were in this order. This is a very common mistake made by those new to firewalls:

  1. Block all traffic
  2. Allow traffic on port 80

In this case, if a connection comes in on port 80 it will be blocked. Even though rule #2 says to allow it, it will be blocked. Lets review why.

This is because rule 1 says to block all traffic. When the request comes in, the firewall looks at its rule. Rule #1 says to block all traffic, so it blocks all traffic. And as the traffic has been blocked, theres nothing else to do. This is an important mistake, because even though there is a later rule that says to allow traffic on port 80, rule #2 will never match anything, because rule 1 matched first, blocked the traffic and processing stopped. Rules process in order, the first match always wins.

This is why firewall rules are called rules. And order always matter. If a request matches a rule, that rule applies as long as no other rule before it applies. Remember, rules are processed in order, 1, 2, 3, 4, etc. When a match occurs, action happens and if that action is to block or allow something, it doesnt matter what any later rules may be, the currently matching rule always "wins".

This is normal behaviour for all firewalls. All firewalls behave this way. Rules are processed in order. Therefore, the order of rules is the most important thing to how a firewall will behave. Make sure you understand the order of your rules to ensure that they are performing as you expect. Most firewall rule problems are due to order.

Stateful packet inspection

When used, the ASL firewall keeps track of the state of network connections. The ASL firewall is designed to distinguish legitimate packets for different types of connections, and to only allow packets matching a known active connection.

This behaviour can be disabled through the rule manager.

Console firewall messages

Note that if you get undesired logging of firewall rule matches to your console, this is not an ASL problem. ASL will not configure your system to do this. If this is happening on your system, this is generally caused by your syslogd or rsyslog configuration.

No known Linux distribution will configure itself to do this by default, so if your system is doing this its either because a third party product has configured your system to do this, or a third party has altered the configuration of your system to do this.

Please contact the third party that has made this modification to your system for assistance.

Fast/Easy Mode

To access the Fast/Easy mode, just click on Configuration Tab in the ASL GUI, then select the "ASL Configuration" menu option. From there scroll down to the Firewall Configuration settings. These are the settings you can configure and what they do:

FW_INBOUND_TCP_SERVICES

List of allowed inbound TCP services, comma delimited. Just provide a list of the ports you want to allow in. Anything not on the list will be denied by default.

Example: 22,21,25,53,80,443,465,110,143,993,995,587,8443,30000

Setting this to "no" will remove all blocks, and will allow all TCP ports to be accessed. Setting this to "" (blank) will block all inbound access to TCP ports.

FW_INBOUND_UDP_SERVICES

List of allowed inbound UDP services, comma delimited. Just provide a list of the ports you want to allow in. Anything not on the list will be denied by default.

Example: 53,123

Setting this to "no" will remove all blocks, and will allow all UDP ports to be accessed. Setting this to "" (blank) will block all inbound access to UDP ports.

FW_OUTPUT_TCP_SERVICES

List of allowed outbound TCP connections, in a comma delimited format. Everything else is blocked, and logged along with the uid for the process that attempted to connect outbound and was denied.

To configure, just provide a list of the ports you want to allow out. Anything not on the list will be denied by default.

You do not need to add any rules for services that are listening. For example, if you allow port 80 in, you do not need to allow port 80 out. ASL uses a stateful firewall and will allow listening services to communicate out. This list is for client initiated requests outbound from the server. For example, if a user wanted to ssh from the server to another server, you would need to allow port 22 out. You do not need to allow port 22 out to allow connections to the ssh service running on the server.

Example: 25,80,443,465,123,53

Remember to allow out other services your system uses such as DNS (port 53). If you do not allow out DNS queries, your system will not be able to resolve any domain names, and things like yum updates, ASL updates, and outbound email will fail.

Setting this to "no" will remove all blocks, and will allow all outbound TCP ports to be accessed. Setting this to "" (blank) will block all outbound access to TCP ports.

FW_OUTPUT_UDP_SERVICES

List of allowed outbound UDP connections, in a comma delimited format. Just provide a list of the ports you want to allow out. Anything not on the list will be denied by default. You do not need to add any rules for services that are listening. For example, if you allow port 53 in, you do not need to allow port 53 out. ASL uses a stateful firewall and will allow listening services to communicate out. This list is for client initiated requests outbound from the server. However, if you had a service that needed to open connections outbound on its own, for example a DNS server would open connections on its own to look up domain names, you would need to allow port 53 out. You do not need to allow port 53 out to allow connections to the dns service running on the server, only if the DNS server needed to make its own independent connections outbound.

Example: 53,123

Also, remember to allow out other services your system uses such as DNS (port 53). If you do not allow out DNS queries, your system will not be able to resolve any domain names, and things like yum updates, ASL updates, and outbound email will fail.

Setting this to "no" will remove all blocks, and will allow all outbound UDP ports to be accessed. Setting this to "" (blank) will block all outbound access to UDP ports.

Engine settings

The Fast/Easy Mode also allows you to configure the core functionality of the netfilter engine. The following is a list of options you can set and what they do.

FW_OUTPUT_MTA

Outbound MTA policy, enabling this rule class will only allow outbound tcp/25 connections for users in the access list /etc/asl/firewall/mta-output-acl. This allows you to restrict outbound SMTP connections to specific users, such as the Mail Servers (MTA) user, which would require your users to use your local mail server for outbound connections. This allows you to filter all outbound email for spam, and prevents your users from setting up spam bots to send email out directly.

Simply put a list of the user names you wish to allow outbound SMTP connections from in the /etc/asl/firewall/mta-output-acl file. The format is one username per line.

In general, you will only want to use the username of the MTA, not your actually users. For example, if qmail-send ran as the user "qmail", you would add that. If postfix ran as the user "postfix", add that. If you had a user "jdoe" that sent mail through your mail server, you would not add jdoe to the /etc/asl/firewall/mta-output-acl file.

So the most important step is to determine what user(s) your mail server uses to send mail. The table below contains a list of the users typically used by the most popular mail servers. Please confirm that this is correct for your mail server configuration, as some vendors change these names:

Mail Server Typical Usernames
Postfix postfix
QMail qmailr
Exim mailnull
Sendmail smmsp

Please check with your mail server vendor to confirm the username that the mail server will use to send mail.

FW_LASSO

Enable/Disable the Spamhaus Lasso blacklist. Enabling this option will block all hosts on the current Spamhaus Lasso blacklist.

Default: No

FW_DSHIELD

Enable/Disable the DShield blacklist. Enabling this option will block all hosts on the current DShield blacklist.

Default: No

FW_TOR

Enable/Disable the TOR exit node list. Enabling this option will block all current TOR exit nodes.

Default: No

FW_PORTSCAN

This will detect portscans against the system and will drop these packets. This will confuse the port scanner, and will cause it to not be able to identify open and closed ports on the system.

Default: yes

FW_BAD_PACKETS

This will detect and drop packets that use bad or invalid options.

Default: yes

FW_SMALL_PACKETS

This will detect packets that are too small to be valid. Each protocol has a minimum size, and this will detect if a packet is too small to be valid.

Default: yes

FW_FRAGMENTS

This will detect fragmented packets. ASL uses a stateful firewall and will automatically handle fragmented packets. This detects if someone is maliciously attempting to bypass the state engine. It is recommend you always have this enabled.

Default: yes

FW_DROP_INVALID

This will detect, and drop all packets with invalid TCP flags, headers or checksums, out of sequence packets which can be caused by sequence prediction or other similar attacks, drop invalid ICMP messages generated by other types of network attacks (Example: recieving a port unreachable ICMP message from a host that the system is not expecting any traffic from) and "orphan" packets that do not belong to a stateful connection.

This setting is important to both stop attacks, and also to prevent false positive shuns under rule 4151. If you connections generate a lot of invalid packets, you will need to enable this so they are not treated as possible attacks.

Default: yes

Note: Do not change the default setting to "no" unless you know what you are doing. This may result in valid connections being shunned. Some poorly configured kernels may not support this capability. All major vendors kernels do this support this capability (Redhat, Centos, ASL), but some vendors have poorly configured their network stack and firewalling capabilities. If you are stuck with one of these inferior kernels, you will need to disable HIDS_4151. We recommend you instead use a more robust kernel, such as the ASL kernel.

FW_DROP_INVALID_LOG

If FW_DROP_INVALID is set to yes, and this option is set to yes, ASL will log all invalid packets it drops.

Default: no

FW_IGNORE_BROADCASTS

If FW_IGNORE_BROADCASTS is set to yes, ASL will silently drop most broadcast packets. ASL will drop the following broadcast packets:

TCP and UDP: 255.255.255.255 Multicast: 224.0.0.0/24

ASL will not drop broadcast packets to the subnets broadcast address.

Default: no

FW_ACCEPT_REDIRECTS

This option controls how the system will respond to redirects. The default setting of "no" is recommended and is also the default setting in Linux. You should not enable this setting unless you know what you are doing. Enabling this can make it possible for an attacker to bypass your firewall rules.

Default: No

FW_ACCEPT_SOURCE_ROUTE

This option controls how the system will respond to a Source Route packet. It is highly recommend you not enable this, and leave the default setting of "no". Source Routed packets can bypass all your firewall rules, are not necessary for proper network functioning and are a relic of IPv4 that should never be used for secure networking.

FW_ICMP_IGNORE_ALL

This option controls how the system will respond to any ICMP broadcast request, such as a ping. It is recommended you not enable this option if you want your system to respond to pings and other ICMP queries.

FW_ICMP_IGNORE_BROADCASTS

This option controls how the system will respond to bogus ICMP broadcast request (so called smurf attacks). It is recommended you enable this option.

FW_IGNORE_ICMP_BOGUS

This option controls how the system will respond to bogus ICMP requests. It is recommended you enable this option.

FW_IPV4_FORWARD

This option enables and disable IPv4 forwarding. If your system is not forwarding traffic to another host, then it is highly recommended you disable forwarding. The default is to disable forwarding.

FW_IPV6_FORWARD

This option enables and disable IPv6 forwarding. If your system is not forwarding traffic to another host, then it is highly recommended you disable forwarding. The default is to disable forwarding.

FW_PROXY_ARP

Enable/Disable proxy arp. It is recommend you not enable this, and leave the default setting of "no". Unless your system is acting as a layer two bridge, you will not need to proxy ARP packets. Enabling this if you are not bridging may result in unexpected networking side effects.

FW_RP_FILTER

Enable/Disable rp_filter. This option controls the Reverse Path Filtering capability in netfilter. This controls whether a response packet is allowed to go out an interface that the request did not originate from. For example, if a system has two interfaces eth0 and eth1, and a packet comes in on eth0. If the system responded, and tried to send the request out eth1 if RP_FILTER was enabled this would be denied. This is to help prevent specific types of routing attacks as well as to restrict egress packets from the system. On systems with complex routing tables, and virtual interfaces this may cause problems as packets will not always exit the same interface they came in on. Therefore, the default is not to enable this. If you do enable this, you should test to make sure your systems configuration will work correctly with this type of restriction.

FW_SYN_COOKIES

Enable/Disable SYN cookies. This option controls the use of SYN cookies. This is a Denial of Service protection feature of netfilter, and we recommend you enable this option.

FW_TCP_ECN

Enable/Disable Excplicit Congestion Notification (ECN). This will allow you to enable/disable ECN.

There are three options:

Yes: Enable ECN

No: Disable ECN

Requested: Only use ECN if the client requests it. (Available in ASL 3.0.26)

For most systems, you should set this to Yes or Requested. We recommend you not disable this unless you know how this will effect your system.

FW_TCP_TIMESTAMPS

Enable/Disable TCP timestamps. This will allow you to enable TCP timestamps which may help with the network performance and quality of some protocols.

For most systems, you should set this to Yes. We recommend you not disable this unless you know how this will effect your system.

FW_TCP_WINDOW_SCALING

Enable/Disable TCP Window Scaling. This will allow you to enable TCP Window scaling which may help with the network performance and quality of some protocols.

For most systems, you should set this to Yes. We recommend you not disable this unless you know how this will effect your system.

Advanced Firewall Rule Manager

The Advanced Rule Manager is designed to let you directly control the netfilter engine in Linux through an advanced GUI system. If you are not familiar with how advanced Linux firewalling works please use the Fast/Easy Mode for simple firewall changes.

Access the firewall rule manager

Step 1: Log into the ASL GUI

Step 2: Click on the Firewall Tab.

Step 3: Click on the firewall rules menu item.

This will open the firewall rules window.

Rule Types

Linux supports three types of firewall rules:

NAT

Network Address Translation Rules: These rules are used to change the source and/or destination of packets.

Mangle

This types of rules are used to modify or "mangle" packets before they are processed

Filter

These types of rules are used to filter or "block" packets. These are the most common rules and are probably the type of rule you want if you want to allow or block traffic to specific port.

Filter subtypes
INPUT

These are rules for traffic going to the server itself. These are the most common subtype, and are what you want if you are looking to allow traffic to a port, or to block traffic to a port.

OUTPUT

These are rules for traffic coming from the server itself. If you have created input rules that are stateful (checking the "stateful" checkbox), then you do not need to create OUTPUT rules for that traffic. The state engine will automatically handle outbound traffic from the server, and will automatically allow this traffic when needed.

These are also different from FORWARD rules. If you are forwarding traffic through your server to a different server, you do not need to create OUTPUT rules for that traffic. OUTPUT rules are only for traffic coming from the server itself. For example, if the server wants to initiate a connection to a mail server on a different server. That would require an OUTPUT rules. If the server was responding to traffic initiated from a different server, and there were stateful INPUT rules to allow that different server to talk to your server on that port, you would not need OUTPUT rules for that traffic.

FORWARD

This subtype is for all all other traffic that is not INPUT or OUTPUT. Specifically, this would be traffic that is passing through the server if the server were acting as a classic firewall for another system. FORWARD rules have nothing to do with the server itself.

Jump Targets

In Linux these tell the rule to jump to an action. This is a way of telling Linux what to do, such as to allow, deny or do something else if the rule is triggered.

ACCEPT

Accept the traffic. This will allow the traffic.

DROP

Silently drop the packet. The client will get no response from the server.

REJECT

Reject the packet. The client will get a reject at the TCP or UDP layer from the server.

Rule options

Using the firewall manager

Do not modify any ASL chains

Do not modify any chains with "ASL" in the name. For example, do not modify the "ASL-Firewall-INPUT" chain. This, and all other ASL-* chains are dynamically modified by ASL on the fly, and changes to these chains may have unpredictable effects and will not be saved by ASL. Your changes will likely be lost, at best, and at worst may cause your system to be unreachable.

If you want to make changes to your firewall, you may modify other chains, or add your own in whatever order you wish.

Adding a firewall rule

Step 1: Select the type of rule you want to add from the list above

Step 2: Select the subtype of the rule you want to add from the list above

Step 3: From the firewall rule manager, select the subtype. For example, to add a filter -> INPUT rule select "INPUT" under the filter tree branch.

this will pull up any INPUT rules you have on the system. To add a new rule select "Add Rule".

Step 4: Select the subtype by selecting it under "Chain". For example, if you want to add an INPUT rule select Table "filter", this should autoselect "INPUT". If it does not just select INPUT as the "Chain".

Step 5: Select the command, Append or Insert. In the Linux firewall, Append means to add the rule after all other rules. Linux firewalling is linear, that means it processes the rules in order "first come first served". Appending places it last. Insert places the rule first. With insert you can also set the rule number in case you prefer that the rule run second, third, etc. Be very careful with Insert as you can end up setting you rules up in the opposite order of what you want (for example, putting your drop rules before your allow rules).

Step 6: Select the protocol by clicking on the "n/a" drop down. The drop down before this allows you to add a "negative", for example if you want to add a rule to say "not tcp". If you want this rule to be for TCP, just select tcp in the drop down, and leave the first drop down as "-".

Step 7: Type in the source IP/Network for this rule. If you want to allow any network, just leave this blank.

Step 8: Type in the destination for this rule. If you are creating a rule for the server itself, for example to allow connections to a specific port, just leave this blank.

Step 9: Type in the source port. If you are a creating a rule to allow connections to the server from another server, leave this blank.

Step 10: Type in the Destination port. For example, if you want to allow connections to port 22, type in 22.

Step 11: Set the interface. If you want to set this rule for all interfaces, leave the default of "n/a".

Step 12: Set the "Jump Target". In the link this is sort of like the "action". You will see a number of options described above, such as ACCEPT, REJECT. Please see above for more information about these. In general, if you want to allow a connection, select "ACCEPT".

Step 13: Select Add Rule.

Your new rule is now implemented.

Examples

Using the Fast/Easy mode

How to block all incoming TCP connections except to certain ports

Step 1: Log into the ASL GUI

Step 2: Click the Configuration Tab

Step 3: Select the ASL Configuration menu option.

Step 4: Scroll down to Firewall configuration

Step 5: Enter the list of TCP ports you want to allow into the FW_INBOUND_TCP_SERVICES list.

Example:

21,22,25,80,110,143,443,993,995,8443,30000

Step 6: Click Update

Your new rule is now implemented.

Using the Advanced Firewall Manager

How to block all incoming connections to a TCP port

Step 1: Log into the ASL GUI

Step 2: Click the Firewall Tab

Step 3: Select the Firewall Rules menu option.

Step 4: Click the "Add Rule" button.

Step 5: For table, select filter.

Step 6: For Chain select "INPUT"

Step 7: For command select "Insert"

Step 6: For protocol select tcp in the drop down, and leave the first drop down as "-".

Step 7: Type in the source IP/Network for this rule. If you want to block all sources, just leave this blank.

Step 8: In the Destination port window type in the port number you wish to block. For example, if you want to block connections to port 22, type in 22. Leave the first drop down as "-".

Step 9: Set the "Jump Target" to "DROP" or "REJECT". DROP will silently block the attempt so the client is not sure why it was blocked, REJECT will send ICMP destination denied packets to the client applications network stack telling the stack that the connection was dropped at the network level.

Step 10: Then press the Add Rule button.

Your new rule is now implemented.

How to restrict all incoming connections to a TCP port except from a specific IP or network

This procedure will allow you to create a ruleset to restrict access to a port, except from specific IPs/networks.

Step 1: Log into the ASL GUI

Step 2: Click the Firewall Tab

Step 3: Select the Firewall Rules menu option.

Step 4: Click the "Add Rule" button.

Step 5: For table, select filter.

Step 6: For Chain select "INPUT"

Step 7: For command select "Insert"

Step 6: For protocol select tcp in the drop down, and leave the first drop down as "-".

Step 7: In the "Source:" field, type in the source IP/Network that you wish to allow access to this port.

Step 8: In the Destination port window type in the port number you wish to restrict access. For example, if you want to restrict connections to port 22, type in 22. Leave the first drop down as "-".

Step 9: Set the "Jump Target" to "ACCEPT", this will allow this IP and/or network to connect to this point.

Step 10: Then press the Add Rule button.

Step 11: Continue adding ACCEPT rules for all the IPs and/or networks you want to allow by repeating Steps 1-10.

Step 12: If you are done adding IPs and/or networks to allow, now you will add in your rule to block all other IPs/Networks. Click the "Add Rule" button.

Step 13: For table, select filter.

Step 14: For Chain select "INPUT"

Step 15: For command select "Insert"

Step 16: For protocol select tcp in the drop down, and leave the first drop down as "-".

Step 17: Type in the source IP/Network for this rule. If you want to block all sources, just leave this blank.

Step 18: In the Destination port window type in the port number you wish to block. For example, if you want to block connections to port 22, type in 22. Leave the first drop down as "-".

Step 19: Set the "Jump Target" to "DROP" or "REJECT". DROP will silently block the attempt so the client is not sure why it was blocked, REJECT will send ICMP destination denied packets to the client applications network stack telling the stack that the connection was dropped at the network level.

Step 20: Then press the Add Rule button.

Your new ruleset to restrict access to a port, except from certain IPs/Networks is now implemented.

Your new rule is now implemented.

How to allow connections to a port

Fast Mode Method

To add TCP ports to the allowed inbound list, please see the FW_INBOUND_TCP_SERVICES option.

To add UDP ports to the allowed inbound list, please see the FW_INBOUND_UDP_SERVICES option.

Advanced Firewall Management Method

Step 1: Log into the ASL GUI

Step 2: Click the Firewall Tab

Step 3: Select the Firewall Rules menu option.

Step 4: Click the "Add Rule" button.

Step 5: For table, select filter.

Step 6: For Chain select "INPUT"

Step 7: For command select "Insert"

Step 6: Select the protocol. For example, for TCP select tcp in the drop down, and leave the first drop down as "-".

Step 7: Type in the source IP/Network for this rule. If you want to allow any network, just leave this blank.

Step 8: In the Destination port window type in the port number you wish to allow. For example, if you want to allow connections to port 22, type in 22. Leave the first drop down as "-".

Step 9: Set the "Jump Target" to "ACCEPT".

How to allow outbound connections to a port

Fast Mode Method

To add TCP ports to the allowed outbound list, please see the FW_OUTPUT_TCP_SERVICES option.

To add UDP ports to the allowed outbound list, please see the FW_OUTPUT_UDP_SERVICES option.


Advanced Firewall Management Method

Note: This is only necessary if you have setup either ASL or your firewall rules to only allow connections outbound to certain ports. By default ASL does not block any outbound ports.

Step 1: Log into the ASL GUI

Step 2: Click the Firewall Tab

Step 3: Select the Firewall Rules menu option.

Step 4: Click the "Add Rule" button.

Step 5: For table, select filter.

Step 6: For Chain select "OUTPUT"

Step 7: For command select "Insert"

Step 6: For protocol select tcp in the drop down, and leave the first drop down as "-".

Step 7: Type in the destination IP/Network for this rule. If you want to allow outbound connections on the port to any network or destination, just leave this blank.

Step 8: In the Destination port window type in the port number you wish to allow. For example, if you want to allow connections outbound port 25, type in 25. Leave the first drop down as "-".

Step 9: Set the "Jump Target" to "ACCEPT".

Step 10: Then press the Add Rule button.

Your new rule is now implemented.

Personal tools