CentOS - Jan 2020 - Blocking attacks from a range of IP addresses

On 1/9/20 2:08 AM, Pete Biggs wrote:
>> Has anyone created a fail2ban filter for this type of attack? As of
>> right now, I have manually banned a range of IP addresses but would
>> like to automate it for the future.
>>
> As far as I can see fail2ban only deals with hosts and not networks - I
> suspect the issue is what is a "network": It may be obvious to
you
> looking at the logs that these are all related, but you run the risk
> that getting denied accesses from, say, 1.0.0.1 and 1.1.0.93 and
> 1.2.0.124 may be interpreted as a concerted attack and you banning half
> the internet - but that may not be a bad thing :-)
>
> What I've done in times of trouble is to be a bit more aggressive in
> why and how hosts are banned. It depends on how you are being attacked,
> but setting the threshold to 1 or 2 failures resulting in a ban and
> then setting the ban time to something fairly short. Repeat offenders
> will then quickly be picked up by the recidive filter and permanently
> banned.
>
> A downside to this is that your firewall filters get very large and
> things will inevitably slow down, but it will at least give you the
> chance to manually block a whole range but still give you a level of
> automated protection.
>
> P.
>
>
> _______________________________________________
> CentOS mailing list
> CentOS at centos.org
> https://lists.centos.org/mailman/listinfo/centos
Since you can configure fail2ban to invoke scripts, I would think it
would be possible to get it to block CIDRs (variable size subnets, i.e.
12.12.0.0/20).? That said, I don't have a quick and easy implementation
on hand.

One thing that has proven useful to me in dealing with these kind of
attacks is to lookup the route for the specified IP address in the
Internet routing tables and then block the entire CIDR block.? You can
also determine which autonomous system (AS number) the attack is coming
from and block other routes owned by that provider.? Other options
include determining the GEOIP location, i.e. country of origin of the
attacks and block by country.

"whois 12.12.12.12" will also provide useful information about the
CIDR
block that you are dealing with.

To look up routes on the Internet, connect to one of the route views
servers, such as this:

$ telnet route-server.ip.att.net

Follow login instructions provided in login banner, then give the command:

show route 12.12.12.12

and it will show you that this IP is part of the larger cidr block
12.12.0.0/20.? So block 12.12.0.0/20 instead of just one IP (assuming
your attacks are coming from a particular block.

You'll have to consider weather the cidr block that you are considering
blocking will impact your users or not.

The att route-server is a juniper.? Some route servers are Cisco's and
there you would use commands like:

show ip route 12.12.12.12

or

show ip bgp 12.12.12.12


An example of a cisco based route server would be to:

telnet route-views.routeviews.org

http://routeviews.org/ lists a bunch of route views servers down near
the bottom, which appear to be accessable by telnet.? Note, quagga based
route servers are open source immitations of the Cisco command interface
and have similar commands to the cisco's.

Nataraj

> > > 
> > As far as I can see fail2ban only deals with hosts and not networks -
I
> > suspect the issue is what is a "network": It may be obvious
to you
> > looking at the logs that these are all related, but you run the risk
> > that getting denied accesses from, say, 1.0.0.1 and 1.1.0.93 and
> > 1.2.0.124 may be interpreted as a concerted attack and you banning
half
> > the internet - but that may not be a bad thing :-)
> > 
> 
> Since you can configure fail2ban to invoke scripts, I would think it
> would be possible to get it to block CIDRs (variable size subnets, i.e.
> 12.12.0.0/20).  That said, I don't have a quick and easy implementation
> on hand.
The OP was looking for an automated way of fail2ban doing it - he had
already sorted out the network range and had stopped this particular
DoS attack. 

P.

Hi!
I usually use a primary ssh jail via f2b, ontop of that I have a reapeat
offenders (usually a check on the f2b logs and rotation needs to be
modified) over a long time.

https://wireflare.com/blog/permanently-ban-repeat-offenders-with-fail2ban/
this could be modified to block bigger pieces of the network. Sadly I have
no direct example for you.

A suggestion is to look into for instance the ipsets from firehol. Unless
you have a more targeted attack using blocklists might be a good option.

Thing is, you might be at a point were any automation does more harm then
good. It depends on what your service does. If it is your homelab with port
22 exposed, the just add big blocks or import firehol-1 and 99% of the
attacks will be dropped. If it is a popular website and you are in need of
blocking webbots then more care needs to be taken.

My suggestion is:

Firehol+change ssh port (if that is the service in question)+ssh
tarpit+repeat offenders

Regards




On Thu, Jan 9, 2020, 20:10 Pete Biggs <pete at biggs.org.uk> wrote:
>
> > > >
> > > As far as I can see fail2ban only deals with hosts and not
networks - I
> > > suspect the issue is what is a "network": It may be
obvious to you
> > > looking at the logs that these are all related, but you run the
risk
> > > that getting denied accesses from, say, 1.0.0.1 and 1.1.0.93 and
> > > 1.2.0.124 may be interpreted as a concerted attack and you
banning half
> > > the internet - but that may not be a bad thing :-)
> > >
> >
> > Since you can configure fail2ban to invoke scripts, I would think it
> > would be possible to get it to block CIDRs (variable size subnets,
i.e.
> > 12.12.0.0/20).  That said, I don't have a quick and easy
implementation
> > on hand.
>
> The OP was looking for an automated way of fail2ban doing it - he had
> already sorted out the network range and had stopped this particular
> DoS attack.
>
> P.
>
>
> _______________________________________________
> CentOS mailing list
> CentOS at centos.org
> https://lists.centos.org/mailman/listinfo/centos
>

On 01/09/2020 02:09 PM, Pete Biggs wrote:
>>> As far as I can see fail2ban only deals with hosts and not networks
- I
>>> suspect the issue is what is a "network": It may be
obvious to you
>>> looking at the logs that these are all related, but you run the
risk
>>> that getting denied accesses from, say, 1.0.0.1 and 1.1.0.93 and
>>> 1.2.0.124 may be interpreted as a concerted attack and you banning
half
>>> the internet - but that may not be a bad thing :-)
>>>
>> Since you can configure fail2ban to invoke scripts, I would think it
>> would be possible to get it to block CIDRs (variable size subnets, i.e.
>> 12.12.0.0/20).  That said, I don't have a quick and easy
implementation
>> on hand.
> The OP was looking for an automated way of fail2ban doing it - he had
> already sorted out the network range and had stopped this particular
> DoS attack. 
>
> P.
>
>
> _______________________________________________
> CentOS mailing list
> CentOS at centos.org
> https://lists.centos.org/mailman/listinfo/centos
Correct. I appreciate all the replies but I used /etc/hosts.deny to deny this
network range of attacks. Again, the reason that fail2ban failed to catch it was
that the attacks were coming from a wide range of subnet addresses and were only
caught by reviewing the log.

It would be nice, however, to have a fail2ban expression that allowed me to
catch the /16 range of addresses needed here.