Mignis is a semantic based tool for firewall configuration. It is designed to help writing iptables rules using an more human-readable syntax, without restricting iptables functionalities.
The traslation from Mignis syntax into the corresponding iptables ruleset has been formally verified in a paper published in Computer Security Foundations Symposium (CSF), 2014:
- Python 2.7 or higher (Python 3.x supported).
Latest working version is available on PyPI:
pip install mignis
Most up-to-date way is to directly install the master branch from GitHub:
pip install --upgrade git+https://github.com/secgroup/Mignis.git
There are many possible ways to write a set of iptables rules for a specific purpose, we decided to use the following approach:
- define each interface and its corresponding subnet.
- bind the interface to the subnet (on the mangle-prerouting chain). This allows us to specify interfaces and/or IP addresses interchangeably (this is for example exploited in the translation of a masquerade rule).
- give higher priority to deny rules, so that they are placed before any other abstract firewall rule (this has to be kept in mind when writing rules).
- use logging rules to analyze mismatched traffic (which will be dropped).
Rules' ordering is fundamental when writing iptables rules. Our approach instead allows to write a set of abstract rules which are order-independent.
usage: mignis.py [--help] [-F] [-c filename] [-w filename | -e | -q query] [-d {0,1,2,3}] [-n] [-f] A semantic based tool for firewall configuration optional arguments: --help, -h show this help message and exit -d {0,1,2,3}, --debug {0,1,2,3} set debugging output level (0-2) -n, --dryrun do not execute/write the rules (dryrun) -f, --force force rule execution or writing possible actions:: -F, --flush flush iptables ruleset -c filename, --config filename read mignis rules from file options for --config/-c: -w filename, --write filename write the rules to file -e, --execute execute the rules without writing to file -q query, --query query perform a query over the configuration (unstable)
Mignis takes a configuration file and generates a series of iptables rules.
Rules can either be written to a file (in a format parsable by the
iptables-restore
command) or directly executed via the iptables
command.
Usage example:
./mignis.py -c config/ex_simple.config -w ex_simple.iptables
This will create an ex_simple.iptables file from the
ex_simple.config configuration. To actually use the rules we just
have to execute iptables-restore ex_simple.iptables
.
OPTIONS default_rules yes logging no INTERFACES lan eth0 10.0.0.0/24 ext eth1 0.0.0.0/0 dummy eth2 none ignore vpn tun0 10.8.0.0/24 ALIASES mypc 10.0.0.2 router_ext_ip 1.2.3.4 malicious_host 5.6.7.8 host_over_vpn 10.8.0.123 remote_host_1 20.20.20.1 remote_host_2 30.30.30.2 remote_host_3 40.40.40.3 remote_hosts (remote_host_1, remote_host_2, remote_host_3) FIREWALL # no restrictions on outgoing connections local > * # ssh accessible from the outside * > local:22 tcp # machines inside the lan are NAT'ed (using masquerade) when communicating through ext lan [.] > ext # forbid the communication with a malicious host lan / malicious_host # dnat to mypc on port 8888 ext > [router_ext_ip:8888] mypc:8888 udp # dnat to host_over_vpn on port 9999 with masquerade ext [.] > [router_ext_ip:9999] host_over_vpn:9999 tcp # allow access to port 80 and 443 on this machine ext > local:(80, 443) tcp # allow only a limited set of hosts to access our vpn remote_hosts > local:1194 udp POLICIES * // * icmp * // * udp * / * CUSTOM # log and accept packets on port 7792 -A INPUT -p tcp --dport 7792 -j LOG --log-prefix "PORT 7792 " -A INPUT -p tcp --dport 7792 -j ACCEPT
Each configuration file needs 6 sections:
OPTIONS: at the moment two generic mignis options can be specified:
default_rules
is used to choose whether to insert default rules or not. Default rules are usually safe to use and are hardcoded into mignis and concern broadcast/multicast packets, invalid packets drops and localhost loopback communication.logging
is used to choose whether to log unexplicitly dropped packets or not (i.e. packets which don't match any rule and get dropped by the default policy).
INTERFACES: defines each interface with their alias (which can be used when writing rules). The syntax is
alias interface-name subnet options
. If the interface doesn't have an ip address the keywordnone
must be used in place of the subnet. At the moment the only option allowed isignore
, which is used to tell mignis to always allow traffic on that interface (i.e. it is not taken into account in firewall rules).ALIASES: defines aliases for IP addresses. The syntax is
alias ip-address
.FIREWALL: contains abstract rules. The syntax is
abstract-rule | iptables-filters
.First we define an address, which is either an interface, an alias or an IP address.
An abstract rule is defined as follows:
from [source_nat] opt [dest_nat] to
- from and to are addresses,
- source_nat is the address from will be SNAT'ed to (it's possible to use "." to indicate a masquerade),
- dest_nat is the address to will be DNAT'ed to,
- opt is one of: "/" (deny with DROP), "//" (deny with REJECT), ">" (one-way forward), "<>" (two-way forward)
Finally an iptables filter is any iptables option used for filtering packets. Common options may be "--icmp-type echo-reply", "-m module", etc.
POLICIES: the default mignis behavior for unmatched packets is to drop them. This section is useful if one wants to reject packets instead, using the mignis syntax for rules matching (only drop or reject rules can be specified). In the example we are rejecting icmp and udp packets, while we're dropping the rest (this last rule may be omitted, we wrote it there only for clarity).
CUSTOM: contains raw iptables rules. Note that you can also modify the tool's behavior here, since you can use the -D and -I switches for deleting and inserting rules in specific locations. We provide this section to add more flexibility, but we cannot guarantee that your custom rules will not conflict with the abstract ones, so please use this section with care and only if you know what you're doing.
Let's see some examples from the configuration above, to clearify how rules can be written and to see how they're translated into iptables rules.
* > local:22 tcp
Allows ssh (tcp port 22) connections towards localhost from any interface.iptables -A INPUT -p tcp --dport 22 -j ACCEPT
lan [.] > ext
Allows packets originating from the lan interface to go to ext, using a source NAT (masquerade).iptables -A FORWARD -i eth0 -o eth1 -j ACCEPT iptables -t nat -A POSTROUTING -s 10.0.0.0/24 -o eth1 -j MASQUERADE
lan / malicious_host
Forbids the communication from the lan towards a malicious host.iptables -A FORWARD -i eth0 -d 5.6.7.8 -j DROP
ext > [router_ext_ip:8888] mypc:8888 udp
UDP packets originating from ext to router_ext_ip on port 8888, are DNAT'ed to mypc on port 8888.iptables -t mangle -A PREROUTING -p udp -i eth1 -d 10.0.0.2 --dport 8888 -m state --state NEW -j DROP iptables -A FORWARD -p udp -i eth1 -d 10.0.0.2 --dport 8888 -j ACCEPT iptables -t nat -A PREROUTING -p udp -i eth1 -d 1.2.3.4 --dport 8888 -j DNAT --to-destination 10.0.0.2:8888
Note: the first mangle rule is used to block packets which are trying to reach mypc bypassing the NAT.
ext [.] > [router_ext_ip:9999] host_over_vpn:9999 tcp
TCP packets originating from ext to router_ext_ip on port 9999, are DNAT'ed to host_over_vpn on port 9999 using a source NAT (masquerade). The masquerade ensures that answers from host_over_vpn are routed through the vpn interface.iptables -t mangle -A PREROUTING -p tcp -i eth1 -d 10.8.0.123 --dport 9999 -m state --state NEW -j DROP iptables -A FORWARD -p tcp -i eth1 -d 10.8.0.123 --dport 9999 -j ACCEPT iptables -t nat -A POSTROUTING -p tcp -s 0.0.0.0/0 -d 10.8.0.123 --dport 9999 -j MASQUERADE iptables -t nat -A PREROUTING -p tcp -i eth1 -d 1.2.3.4 --dport 9999 -j DNAT --to-destination 10.8.0.123:9999
Note: the first mangle rule is used to block packets which are trying to reach host_over_vpn bypassing the NAT.
ext > local:(80, 443) tcp
Allow access from ext to port 80 and 443 on the local machine.iptables -A INPUT -p tcp -i eth1 --dport 80 -j ACCEPT iptables -A INPUT -p tcp -i eth1 --dport 443 -j ACCEPT
remote_hosts > local:1194 udp
Only the list of hosts specified in remote_hosts can connect to our VPN.iptables -A INPUT -p udp -s 20.20.20.1 --dport 1194 -j ACCEPT iptables -A INPUT -p udp -s 30.30.30.2 --dport 1194 -j ACCEPT iptables -A INPUT -p udp -s 40.40.40.3 --dport 1194 -j ACCEPT
- Lists operations for excluding hosts/lists from a list. For example
if we define a list alias "list1 (eth0, eth1)" and want a rule that
is valid for list1 except for the host 1.1.1.1 (which belongs to
the interface eth0), we can write
list1/1.1.1.1 > eth2
. - Improving checks for identifying overlapping rules.
- Rules queries to list all the connections that match a particular host, this is useful to see all the packets a host can send/receive. This has to be expanded with lists and rules (exploiting the overlapping checks).
- Complete code rewrite with a modular compiler-like design.
- Support multiple firewall languages (iptables, nftables, Cisco, etc.)
- Abstract-level rules optimizations.
- Accept different kinds of configuration files (e.g. JSON, python scripts) and/or consider a richer language for writing the rules.
- Provide a 2nd-level abstract semantic using security roles.