Oct 3, 201810 min read

Protect Your Mac with PF, the All Powerful Firewall

PF

This post is material that comes from a slightly earlier version but is very relevant none the less. However, you should be aware that an OS Update will wipe out your existing pf.conf ... very bad, because you won't notice. What you should do is copy your pf.conf into a file that won't be destroyed, for example, your domain-name.pf.conf and then, if your plist file, use that file name, not pf.conf. This Mac is a Mac Mini 2012. July. It is NOT running Apple's Server App. It IS running all the components of a server that I have either built myself or added with Homebrew. These include the following. Apache 2.4, Postfix, Dovecot, Spamassassin, Amavis and numerous other small binaries needed to support them, and other things I'm doing. This set of configurations are valid for OSX though, up to at least the latest macOS High Sierra Version 10.13.6. When the new release appears, I'll check and update this post. That is, the sample pf.conf that follows this "tutorial" which in some cases is inaccurate... but mine works. It contains IPv6 settings as well as IPv4, and some settings that I'm working on as an experiment. Remember, pf won't work even if you load it, unless it's Enabled. See below, and in the sample pf.conf. Don't forget to read the onboard man pages. $man pf.conf and $man pfctl

The Main Description Mac OS X 10.6 (and earlier) came with IPFW, a port of FreeBSD's stateful firewall<1>. IPFW was deprecated in OS X 10.7, and was completely removed in OS X 10.10; it was replaced with PF. PF (Packet Filter) is OpenBSD's system for filtering TCP/IP traffic and doing Network Address Translation<2>. PF in OS X, however, appears to be based on the FreeBSD port of PF<3>, but with some notable additions (see below). Like FreeBSD 9.X and later, OS X appears to use the same version of PF as OpenBSD 4.5. Note that the latest OpenBSD version is 5.6 (as of January 2015);and the configuration syntax for PF changed around 4.6/4.7. Apple has enhanced PF so that various system components might choose to enable and disable PF, as indicated by the following snippet in /etc/pf.conf:

# This file contains the main ruleset, which gets automatically loaded
# at startup.  PF will not be automatically enabled, however.  Instead,
# each component which utilizes PF is responsible for enabling and disabling
# PF via -E and -X as documented in pfctl(8).  That will ensure that PF
# is disabled only when the last enable reference is released.

These two flags, -E and -X, are absent from pfctl on other BSDs. Here's how they are documented in pfctl(8):

     -E      Enable the packet filter and increment the pf enable reference count.
     -X token
             Release the pf enable reference represented by the token passed.
     -s References  Show pf-enable reference statistics (pid/name of enabler, token, timestamp).

The main PF configuration file is /etc/pf.conf, which defines the following main ruleset by default in OS X 10.9 & 10.10:

scrub-anchor "com.apple/*"
nat-anchor "com.apple/*"
rdr-anchor "com.apple/*"
dummynet-anchor "com.apple/*"
anchor "com.apple/*"
load anchor "com.apple" from "/etc/pf.anchors/com.apple"

The main ruleset loads sub rulesets defined in /etc/pf.anchors/com.apple, using anchor<4>:

anchor "200.AirDrop/*"
anchor "250.ApplicationFirewall/*"

The launchd configuration file for PF is /System/Library/LaunchDaemons/com.apple.pfctl.plist. PF is disabled by default:

$ sudo pfctl -s info
No ALTQ support in kernel
ALTQ related functions disabled
Status: Disabled                              Debug: Urgent

Application Firewall

OS X v10.5.1 and later include Application Firewall that allow the users to control connections on a per-application basis (rather than a per-port basis)<5>. Application Firewall is disabled by default. After enabling the Application Firewall (System Preferences -> Security & Privacy -> Firewall -> Turn On Firewall), you'll find PF is enabled too:

$ sudo pfctl -s info
Status: Enabled for 0 days 00:02:03           Debug: Urgent

$ sudo pfctl -s References
TOKENS:
PID      Process Name                 TOKEN                    TIMESTAMP
618      socketfilterfw               9813589183660731843      0 days 00:03:31

$ sudo pfctl -a com.apple -s rules
anchor "200.AirDrop/*" all
anchor "250.ApplicationFirewall/*" all

Apparently Application Firewall enables PF using pfctl -E. In addition to its own rules, Application Firewall generates a set of dynamic rules (sub ruleset) for PF through anchor point com.apple/250.ApplicationFirewall. At this stage, the sub ruleset is empty, which got someone confused.

But if either "Enable stealth mode" or "Block all incoming connections" is checked in Firewall Options..., dynamic rules for PF will indeed be created:

$ sudo pfctl -a com.apple/250.ApplicationFirewall -s rules
scrub in all fragment reassemble
block drop in inet proto icmp all icmp-type echoreq
block drop in inet proto icmp all icmp-type echoreq
block drop in inet6 proto ipv6-icmp all icmp6-type echoreq
Note there is a bug in Apple's implementation of PF! According to pfctl(8):

If the anchor name is terminated with a `*' character, the -s flag will recursively print all anchors in a brace delimited block. but it produces an error instead:

$ sudo pfctl -a 'com.apple/*' -sr
anchor "*" all {
pfctl: DIOCGETRULES: Invalid argument
}
anchor "*" all {
pfctl: DIOCGETRULES: Invalid argument
}

We have to use the full anchor path:

$ sudo pfctl -v -s Anchors
  com.apple
  com.apple/200.AirDrop
  com.apple/200.AirDrop/Bonjour
  com.apple/250.ApplicationFirewall

$ sudo pfctl -a "com.apple/200.AirDrop/Bonjour" -sr
pass in on p2p0 inet6 proto udp from any to any port = 5353 keep state
pass out on p2p0 proto tcp all flags any keep state

As you can see, a set of dynamic PF rules is created for AirDrop too. I surmise they are still created by Application Firewall, because according to the output of pfctl -s References, PF has only been enabled once, by Application Firewall.

Command Line Besides using the Security & Privacy Preference pane, you can also configure the Application Firewall from the command line. The utilities for Application Firewall are stored in /usr/libexec/ApplicationFirewall. The default configuration file is /usr/libexec/ApplicationFirewall/com.apple.alf.plist; and the running configuration file is /Library/Preferences/com.apple.alf.plist<6>. ▸ Stopping and starting Application Firewall is easy enough, using launchd<7>. To stop:

$ sudo launchctl unload /System/Library/LaunchAgents/com.apple.alf.useragent.plist 
$ sudo launchctl unload /System/Library/LaunchDaemons/com.apple.alf.agent.plist 
▸ To start:

$ sudo launchctl load /System/Library/LaunchDaemons/com.apple.alf.agent.plist 
$ sudo launchctl load /System/Library/LaunchAgents/com.apple.alf.useragent.plist 

We can configure the settings of Application Firewall using socketfilterfw:

usage: /usr/libexec/ApplicationFirewall/socketfilterfw <-c> <-w> <-d> <-l> <-t> <-u> <-b> <-l> \
    <-a listen or accept> <-p pid to write> <--getglobalstate> <--setglobalstate on | off> \
    <--getblockall> <--setblockall on | off> <--listapps> \
    <--getappblocked <path>> <--blockapp <path>> <--unblockapp <path>> \
    <--add <path>> <--remove <path>> <--getallowsigned> <--setallowsigned> \
    <--getstealthmode> <--setstealthmode on | off> \
    <--getloggingmode> <--setloggingmode on | off> \
    <--getloggingopt> <--setloggingopt throttled | brief | detail>

pflog

Logging support for PF is provided by pflog. The pflog interface is a pseudo-device which makes visible all packets logged by PF. Logged packets can easily be monitored in real time by invoking tcpdump on the pflog interface. ▸ Create a pflog interface:

$ man 4 pflog
$ sudo ifconfig pflog0 create
▸ Monitor all packets logged by PF:

$ sudo tcpdump -n -e -ttt -i pflog0
▸ Destroy the pflog interface when done:

$ sudo ifconfig pflog0 destroy

precedence

If two firewalls, Application Firewall & PF, are running, you may wonder whose rules take precedence. Let's find out. The logs of Application Firewall are saved in /var/log/appfirewall.log. You'll see a lot entries like the following, repeating roughly 2 times per minute on my iMac:

Jan 20 00:03:35 manjusri.local socketfilterfw<228> <Info>: Dropbox109: Deny UDP CONNECT (in:22 out:0)
Jan 20 00:03:35 manjusri.local socketfilterfw<228> <Info>: ntpd: Deny UDP CONNECT (in:2 out:0)
Jan 20 00:03:35 manjusri.local socketfilterfw<228> <Info>: netbiosd: Deny UDP CONNECT (in:13 out:0)
Jan 20 00:03:44 manjusri.local socketfilterfw<228> <Info>: Stealth Mode connection attempt to UDP 3 time
Jan 20 00:03:44 manjusri.local socketfilterfw<228> <Info>: Stealth Mode connection attempt to TCP 2 time

Add the following as the first rule of /etc/pf.conf:

set skip on lo0

Add the following 3 lines to /etc/pf.conf (to block incoming traffic but allow outgoing traffic):

pass in quick proto udp to any port 5353
block in
pass out quick

The first rule is to allow incoming Bonjour traffic. In a hostile environment, e.g., a public WiFi, we'll put the above 3 lines at the end of the file to block all incoming traffic, in which case, the sub rulesets in anchor "com.apple" will have no effect! Note For each packet or connection evaluated by PF, the last matching rule in the ruleset is the one which is applied. In work environment, you can put the 3 lines right above the line:

anchor "com.apple/*"

Reload /etc/pf.conf:

$ sudo pfctl -f /etc/pf.conf

Show the currently loaded filter rules:

$ sudo pfctl -s rules
scrub-anchor "com.apple/*" all fragment reassemble
block drop in all
pass out all flags S/SA keep state
anchor "com.apple/*" all

Check /var/log/appfirewall.log again. You'll find no new log entry for Application Firewall appears in the file.

So one can conclude that PF rules are applied first, then the rules for Application Firewall.

SSH

To enable OpenSSH server on OS X, in the Sharing Preference pane of System Preferences, check "Remote Login". Or from the command line:

$ sudo launchctl load -w /System/Library/LaunchDaemons/ssh.plist

launchctl(1) says such about the -w flag:

-w Overrides the Disabled key and sets it to false. In previous versions, this option would modify the configuration file. Now the state of the Disabled key is stored elsewhere on-disk. but where exactly is the 'elsewhere'? After some digging, I find it is /private/var/db/launchd.db/com.apple.launchd/overrides.plist. However, I don't like the default configuration for sshd. I prefer to have password authentication disabled. Add the following options to /etc/sshd_config:

PermitRootLogin no
PasswordAuthentication no
ChallengeResponseAuthentication no

Restart sshd:

$ sudo launchctl unload /System/Library/LaunchDaemons/ssh.plist
$ sudo launchctl load /System/Library/LaunchDaemons/ssh.plist
Note to allow incoming traffics to the OpenSSH server through Application Firewall, you must allow incoming connections for /usr/libexec/sshd-keygen-wrapper, either in System Preferences -> Security & Privacy -> Firewall -> Firewall Options..., or from the command line:

$ sudo /usr/libexec/ApplicationFirewall/socketfilterfw --add /usr/libexec/sshd-keygen-wrapper

Configuring PF

The Application Firewall's rule of allowing all incoming incoming traffics to the OpenSSH server offers no defense against brute force attack. Leaving the ssh port open on the internet, the server will get thousands of brute force login attempts each day. PF provides an elegant solution to this problem. ▸ Append the following lines to /etc/pf.conf (see Section 30.3.3.5. Using Overload Tables to Protect SSH of FreeBSD Handbook for an explanation):

 
table <bruteforce> persist
block quick from <bruteforce>
pass in inet proto tcp to any port ssh \
    flags S/SA keep state \
    (max-src-conn 5, max-src-conn-rate 5/5, \
     overload <bruteforce> flush global)
▸ Reload /etc/pf.conf:

$ sudo pfctl -f /etc/pf.conf

Over time, the table bruteforce will be filled by overload rules and its size will grow incrementally, taking up more memory. We can expire table entries using pfctl. For example, this command will remove bruteforce table entries which have not been referenced for a day (86400 seconds):

$ sudo pfctl -t bruteforce -T expire 86400

To automate the process, let's create a timed job using launchd that runs the above command once per day<8>.

▸ Create a launchd configuration file /Library/LaunchDaemons/edu.ucsc.manjusri.pfctl-expire.plist, with the following content:

<?xml version="1.0" encoding="UTF-8"?><!DOCTYPE plist PUBLIC "-//Apple//DTD PLIST 1.0//EN" "http://www.apple.com/DTDs/PropertyList-1.0.dtd"><plist version="1.0"><dict> <key>Label</key> <string>edu.ucsc.manjusri.pfctl-expire</string> <key>WorkingDirectory</key> <string>/var/run</string> <key>UserName</key> <string>root</string> <key>GroupName</key> <string>wheel</string> <key>Program</key> <string>/sbin/pfctl</string> <key>ProgramArguments</key> <array> <string>pfctl</string> <string>-t</string> <string>bruteforce</string> <string>-T</string> <string>expire</string> <string>86400</string> </array> <key>StartCalendarInterval</key> <dict> <key>Hour</key> <integer>10</integer> <key>Minute</key> <integer>10</integer> </dict></dict></plist>▸ Start the timed job:

$ sudo launchctl load /Library/LaunchDaemons/edu.ucsc.manjusri.pfctl-expire.plist
P.S. There are a few articles on the Internet on using PF on Mac OS X, but they often bypass the configuration file /etc/pf.conf<9>. If one takes that route, one must disable the Application Firewall. Otherwise Application Firewall will enable PF using the ruleset in /etc/pf.conf. Only one ruleset will get loaded at last and become effective; but which one wins will probably be indeterministic or at least could be a surprise. I choose the approach described in this article, because:

I alway like to try something different
I prefer layered defense. In this case, I have 2 firewalls running on the Mac.

References

Content is available under Creative Commons Attribution 4.0 International Public License unless otherwise noted.
Privacy

Sample pf.conf called by my .plist

# sample firewall rules
# to enable:
# pfctl -e -f chalmers.pf.conf
# to disable:
# pfctl -d
# To check syntax. sudo pfctl -vnf /etc/chalmers.pf.conf
set block-policy drop
set fingerprints "/etc/pf.os"
set ruleset-optimization basic
set skip on lo0

# Normalization
# Scrub incoming packets
scrub in all no-df

#
# com.apple anchor point
#
pflog_logfile="/var/log/pflog"
scrub-anchor "com.apple/*"
nat-anchor "com.apple/*"
rdr-anchor "com.apple/*"
dummynet-anchor "com.apple/*"
anchor "com.apple/*"
load anchor "com.apple" from "/etc/pf.anchors/com.apple"

anchor "emerging-threats"
load anchor "emerging-threats" from "/etc/pf.anchors/emerging-threats"

### START Custom Rules ###

AllowedIn = "{ 192.168.0.0/24, 192.168.1.0/24 }"

table  persist file "/etc/badguys1" file "/etc/badguys2"

# to add a host to the bad list
# pfctl -t badhosts -Tadd 
# to see the current bad hosts 
# pfctl -t badhosts -T show
# table  persist
# block the jerks
block on en1 from  to any
block on en0 from  to any

table  persist


# Block by default
#block in log

# Block to/from illegal destinations or sources
block in log quick from no-route to any

# Allow critical system traffic
pass in quick inet proto udp from any port 67 to any port 68

# Allow ICMP from Mac Mini LAN
#pass in log proto icmp from 66.209.67.0/24

# Allow all lan traffic in (if you are on a LAN that is yours)
#pass in from 10.0.0.0/24 to any
# Allow all VPN traffic in (if there is a VPN LAN)
#pass in from 10.0.1.0/24 to any

# allow good places everything, if you want
pass in from {} to any

pass in inet proto tcp from  to any
pass in inet proto udp from  to any

# Allow outgoing traffic
pass out proto tcp from any to any keep state
pass out proto udp from any to any keep state

# allow stuff you want
pass in proto tcp from any to any port 80
pass in proto tcp from any to any port 443
pass in proto tcp from any to any port 21
pass in proto tcp from any to any port 2121
pass in proto tcp from any to any port 22
pass in proto tcp from any to any port 110
pass in proto tcp from any to any port 143
pass in proto tcp from any to any port 465
pass in proto tcp from any to any port 587
pass in proto tcp from any to any port 993
pass in proto tcp from any to any port 995

# block the jerks
block on en1 from {} to any
block on en0 from {} to any


pass proto ipv6-icmp from any to any


# Additional IPv6 settings.

ext_if="en0"

icmp6_types="{ 2, 128 }" # packet too big, echo request (ping6)
# Neighbor Discovery Protocol (NDP) (types 133-137):
#   Router Solicitation (RS), Router Advertisement (RA)
#   Neighbor Solicitation (NS), Neighbor Advertisement (NA)
#   Route Redirection
icmp6_types_ext_if="{ 128, 133, 134, 135, 136, 137 }"
tcp46_services="{ 22 }" # ssh; we allow this in for ALL machines for TCP<46>
tcp46_services_ext_if="{ 53 }" # DNS zone transfer
udp6_services_ext_if="{ 53, 123, 1194, 546}" # 546 == dhcpv6-client

# IPv6  ... These don't work.
#pass in quick on $ext_if inet6 proto ipv6-icmp icmp6-type $icmp6_types keep state

#pass in quick on $ext_if inet6 proto ipv6-icmp from any to
	#{ ($ext_if ), ff02::1/16 } icmp6-type $icmp6_types_ext_if keep state
#pass in quick on $ext_if inet6 proto tcp from any to any port
	#$tcp46_services flags S/SA keep state
#pass in quick on $ext_if inet6 proto tcp from any to
	#( $ext_if ) port $tcp46_services_ext_if flags S/SA
	#keep state
#pass in quick on $ext_if inet6 proto udp from any to
	#( $ext_if ) port $udp6_services_ext_if keep state


table  persist
block quick from 
pass in inet proto tcp to any port ssh \
    flags S/SA keep state \
    (max-src-conn 5, max-src-conn-rate 5/5, \
     overload  flush global)

### END Custom Rules ###

Affinity Inspiration