Re: rate limit SIP INVITES

[sean darcy <seandarcy2@gmail.com>]

On 9/27/20 4:59 PM, Pablo Neira Ayuso wrote:
> On Sun, Sep 27, 2020 at 11:42:08AM -0400, sean darcy wrote:
>> On 9/27/20 10:03 AM, Pablo Neira Ayuso wrote:
>>> On Sun, Sep 27, 2020 at 03:54:47PM +0200, Pablo Neira Ayuso wrote:
>>>> On Sat, Sep 26, 2020 at 03:10:24PM -0400, sean darcy wrote:
>>>>> nftables-0.9.6
>>>>>
>>>>> I'm running a VOIP server. There are lots of script kiddies who will bang
>>>>> away with 10/sec SIP INVITES or REGISTERS .
>>>>>
>>>>> In iptables you can match on the string:
>>>>>
>>>>> -A SIP   -i eth0 -p udp -m udp --dport 5060 -m string   --string "INVITE"
>>>>> --algo bm --from 23 --to 28 -m comment --comment "Catch SIP INVITEs" -j
>>>>> SIPINVITE
>>>>>
>>>>> -A SIP   -i eth0 -p udp -m udp --dport 5060 -m string   --string "REGISTER"
>>>>> --algo bm --from 23 --to 30 -m comment --comment "Catch SIP REGISTERs" -j
>>>>> SIPREGISTER
>>>>>
>>>>> I'm looking at RAW to do the same:
>>>>
>>>> nft add rule x y udp dport 5060 @th,64,48 0x494e56495445 counter
>>>>
>>>> @th => transport header
>>>> 64  => from bit number 64 (8 bytes after the UDP header)
>>>> 48  => extract 48 bits (6 bytes for INVITE)
>>>
>>> @th,offset,length
>>>
>>> where offset and length are expressed in bits.
>>>
>> Thanks for the response.
>>
>> I corrected it , but it didn't work:
>>
>> nft list chain filter raw
>> table ip filter {
>> 	chain raw {
>> 		type filter hook prerouting priority raw; policy accept;
>> 		udp dport 5060 @th,184,48 80600803923013 counter packets 0 bytes 0
>> 		udp dport 5060 @th,184,64 5928222864759342418 counter packets 0 bytes 0
> 
> This should be:
>                                 @th,64,48 0x494e56495445 counter
> 
> you specify offset to 184, that does not look fine.
> 
> If you want to match INVITE right after the UDP header, in the initial
> 6 bytes of the payload, then offset is 64 bits give that UDP header is
> 8 bytes (64 bits).
> 
> Note that @th specifies that the offset is relative to the transport
> header offset. Similarly, @nh specifies the offset relative to the
> network header.
> 
> I tried it here with nc -u and sending the string INVITE and it works
> fine.
> 

It's working. Thanks for all the quick and responsive help.

For those who find this exchange, and are as clueless as I am about bit 
counting network packets, and the differences between link layer, 
network header, and transport header:

table ip filter {
	chain raw {
		type filter hook prerouting priority raw; policy accept;
		udp dport 5060 @th,64,48 80600803923013 counter packets 221 bytes 162020
		udp dport 5060 @th,64,64 5928222864759342418 counter packets 67 bytes 
39671
		udp dport 5060 @ll,336,48 80600803923013 counter packets 221 bytes 162020
		udp dport 5060 @ll,336,64 5928222864759342418 counter packets 67 bytes 
39671
		udp dport 5060 @nh,224,48 80600803923013 counter packets 221 bytes 
162020 drop
		udp dport 5060 @nh,224,64 5928222864759342418 counter packets 67 bytes 
39671 drop
	}
}

I repeat my suggestion that "nft list" show the pattern to be matched in 
hex, as it is in the command.

sean