nftables-init.rules

#                                  
#   Netfilter's NFTable firewall   
#                                  
# This is just a ruleset to play around with the syntax introduced
# in nftables and itis my way of getting to know it.
#
# Here might be dragons!
#
# To invoke:
#
#   $ sudo iptable-save > iptables.backup
#   $ sudo iptables -P INPUT DROP
#   $ sudo iptables -F
#   $ sudo iptables -X
#   $ sudo nft flush ruleset && sudo nft -f nftables-init.rules
#
# To get back to your iptables ruleset:
#
#  $ sudo nft flush ruleset
#  $ sudo iptables-restore < iptables.backup
#
#  BEWARE: during the above commands there is a short moment where 
#          there are no firewall rules active. That is why the default
#          policy is changed to drop all traffic. But still you 
#          should make sure to only try this on trusted networks! 
#

flush ruleset

define admin = { 12.34.56.78/29, 10.11.12.0/8, 172.16.1.0/16 }
define google_dns = { 8.8.8.8, 8.8.4.4 }
define mailout = { 127.0.0.1 }

table nat {
    chain prerouting {
        type nat hook prerouting priority 0
        
        # initiate redirecting on the local machine and redirect incoming 
        # traffic on port 21 to 21212 which is nice for docker for example
        redirect
        tcp dport 21 redirect to 21212
    }

    chain postrouting {
        type nat hook postrouting priority 0
        
        # we need this chain even if there are no rules for the return 
        # path otherwise the path will not exist
    }
}

table inet filter {
    chain input {
        type filter hook input priority 0; policy accept

        # drop all bad actors before we do rel/est
        ip saddr @blackhole drop
        
        # connection track and accept previous accepted traffic
        ct state established,related accept
        
        # localhost godmode
        iif lo accept
        
        # if the connection is NEW and is not SYN then drop
        tcp flags != syn ct state new log prefix "FIRST PACKET IS NOT SYN" drop

        # new and sending FIN the connection? DROP!
        tcp flags & (fin|syn) == (fin|syn) log prefix "SCANNER1" drop

        # i don't think we've met but you're sending a reset?
        tcp flags & (syn|rst) == (syn|rst) log prefix "SCANNER2" drop

        # 0 attack? 
        tcp flags & (fin|syn|rst|psh|ack|urg) < (fin) log prefix "SCANNER3" drop 

        # xmas attack. lights up everything
        tcp flags & (fin|syn|rst|psh|ack|urg) == (fin|psh|urg) log prefix "SCANNER4" drop

        # if the ctstate is invalid 
        ct state invalid log flags all prefix "Invalid conntrack state: " counter drop

        # open ssh, http and https and give them the new state 
        tcp dport { ssh, http, https } ct state new accept

        # handle packets from iprange to admin chain
        ip saddr $admin jump admin
        
        # icmpv6 for ipv6 connections
        ip6 nexthdr icmpv6 icmpv6 type {
            destination-unreachable, packet-too-big, time-exceeded,
            parameter-problem, nd-router-advert, nd-neighbor-solicit,
            nd-neighbor-advert
        } limit rate 100/second accept

        # icmp for ipv4 connections 
        ip protocol icmp icmp type {
            destination-unreachable, router-advertisement, 
            time-exceeded, parameter-problem 
        } limit rate 100/second accept

        # otherwise we drop, drop, drop
        #
        # when you are troubleshooting uncomment the next line.
        # log prefix "Incoming packet dropped: "
    }
    
    chain output {
        type filter hook output priority 0; policy accept

        # connection track and accept previous accepted traffic
        ct state established,related accept
        
        # all powerfull... as long as it is to localhost
        oif lo accept
       
        # allow DNS request if they are not to Google's DNS
        # i think this would qualify as torture, but I
        # have never claimed this set to be technically 
        # or morraly sound. 
        udp dport 53 ip daddr $google_dns accept
        tcp dport 53 ip daddr $google_dns accept
	
        # allow dhcp
        udp dport 67 accept

        # youtube needs this for tracking where you are in the video... weird. 
        udp dport 443 accept

        # mail, really? are you malwa... -uhm- mailware!
        tcp dport {25,465,587} ip daddr != $mailout log prefix "SPAMALERT!" drop

        # allow web requests
        tcp dport { http, https } ct state new accept

        # limit outgoing icmp type 8 traffic
        ip protocol icmp icmp type echo-request limit rate 1/second log accept

        # log packet before it is dropped
        log flags all prefix "Outgoing packet dropped: "
    }
    
    chain admin {
        tcp dport ssh ct state new log prefix "Admin connection:" accept
    }
    
    set blackhole {
    	# to add ip's to the blacklist you could use the commandline _nft_ tool ie:
	# nft add element ip filter blackhole { 192.168.1.4, 192.168.1.5 }
	
        # blackhole ipset where we set the type of element as ipv4
        type ipv4_addr
        
        # we will set a timer on the element after which it is cleared
        flags timeout

        # the value of the timer
        timeout 1d
    }
}