<html>
<head>
<meta http-equiv="content-type" content="text/html; charset=UTF-8">
</head>
<body>
<font face="Tahoma">Hi all.<br>
<br>
I'm not one to b**ch & moan in public, but per subject, I
could not let this one slide.<br>
<br>
And if you get this from multiple mailing lists, apologies for
that - I'm just trying to make sure that this reaches as wide an
audience as possible, on the continent.<br>
<br>
SEACOM (AS37100) acquired MacroLan (AS37353) a couple of years
ago. MacroLan is now part of the SEACOM family, and while we are
in the process of integrating that network into AS37100, some
existing services continue to be delivered on AS37353 until that
exercise is completed.<br>
<br>
One of the customers that AS37353 was providing services to was
Cloud Innovation, in Cape Town. From a routing perspective,
because Cloud Innovation had no AS number for this service, all of
their IP address space was being originated by AS37353, on their
behalf.<br>
<br>
In December of 2019, AS37353 ceased providing services to Cloud
Innovation. That is 6 months ago.<br>
<br>
In recent days, it came to SEACOM's attention that some of Cloud
Innovation's IP address space was being originated by AS37353 -
specifically, 156.241.3.0/24 - even though we were sure that they
were no longer a customer of AS37353 since December of 2019. At
first, we thought this was a cached entry in a number of popular
looking glasses, but then every looking glass had the same entry,
which made this an unlikely bug.<br>
<br>
As of yesterday afternoon, see below what Telia's looking glass
was saying about this prefix:<br>
<br>
*****<br>
<br>
Path #1: Received by speaker 0<br>
4809 134190 37353<br>
2.255.249.42 (metric 2134) from 2.255.253.101 (80.91.242.40)<br>
Origin incomplete, localpref 200, valid, internal, best,
group-best, import-candidate<br>
Communities:<br>
<br>
1299:431<br>
(RPKI state Unknown)<br>
<br>
1299:1000 1299:30000 1299:30600 23456:20413 45352:4500 45352:9204<br>
<br>
*****<br>
<br>
But when I run a traceroute from my house to that prefix, it
clearly was not ending up in Cape Town, where AS37353's main
operation resides:<br>
<br>
*****<br>
<br>
MacBook-Pro-7:~ tinka$ traceroute -I 156.241.3.1<br>
traceroute to 156.241.3.1 (156.241.3.1), 64 hops max, 72 byte
packets<br>
1 172.16.0.254 (172.16.0.254) 14.824 ms 11.522 ms 3.525 ms<br>
2 xe-1-3-0-1064.er-01-jnb.za.seacomnet.com (105.22.37.13) 5.620
ms 9.714 ms 9.887 ms<br>
3 ce-0-2-0-0.cr-02-jnb.za.seacomnet.com (105.16.28.2) 175.232
ms 172.699 ms 175.896 ms<br>
4 xe-0-0-0-8.cr-02-cpt.za.seacomnet.com (105.16.9.182) 164.496
ms 163.578 ms 163.546 ms<br>
5 105.16.14.153 (105.16.14.153) 169.812 ms 171.272 ms 177.115
ms<br>
6 xe-0-0-0-0.br-02-lhr.uk.seacomnet.com (105.16.34.253) 168.911
ms 172.958 ms 165.165 ms<br>
7 82.112.115.169 (82.112.115.169) 172.700 ms 176.482 ms
174.375 ms<br>
8 ae-17.r05.londen12.uk.bb.gin.ntt.net (129.250.2.147) 672.099
ms 613.617 ms 615.109 ms<br>
9 ae-2.r24.londen12.uk.bb.gin.ntt.net (129.250.4.244) 181.952
ms 183.087 ms 187.302 ms<br>
10 ae-16.r20.frnkge13.de.bb.gin.ntt.net (129.250.3.13) 190.511
ms 185.579 ms 187.058 ms<br>
11 ae-3.r20.sngpsi07.sg.bb.gin.ntt.net (129.250.4.17) 520.882
ms 613.982 ms 615.275 ms<br>
12 ae-9.r24.tkokhk01.hk.bb.gin.ntt.net (129.250.7.67) 612.301
ms 586.886 ms 436.711 ms<br>
13 ae-1.r03.tkokhk01.hk.bb.gin.ntt.net (129.250.6.98) 614.470
ms 613.416 ms 614.281 ms<br>
14 ce-0-3-0-3.r03.tkokhk01.hk.ce.gin.ntt.net (203.131.241.126)
614.128 ms 613.661 ms 615.416 ms<br>
15 * * *<br>
16 * * *<br>
17 156.241.3.1 (156.241.3.1) 494.400 ms 410.180 ms *<br>
MacBook-Pro-7:~ tinka$<br>
<br>
*****<br>
<br>
So we, then, realized that this was a fraudulent use of MacroLan's
AS37353, to which we had given no express permission.<br>
<br>
As luck would have it, due to my days living and working in
Malaysia, I know the good folk that operate AS134190 (IPDC
Solutions), who was the upstream providing transit for this
prefix. So I rang them up yesterday afternoon, told them what was
happening, and within the hour, they got that eBGP session
shutdown. I am most grateful to them for their quick response and
immediate understanding of what was going on. Even with all the
technology we have today, it, many times, comes down to having
good friends in good places.<br>
<br>
Anyway, it turns out the ISP that had acquired this prefix from
Cloud Innovation is based in Manila, Philippines. When IPDC
delivered their transit service to them in Manila, that ISP
informed them that they should use AS37353 for the eBGP session.
Yes, one could argue that IPDC should have done their checks to
ensure that the AS number being provided by their customer belongs
to them, but to be honest, I'm not too bothered about that
compared to Cloud Innovation's thinking that it was okay to use
another network's AS number in order to deliver their services to
their customers.<br>
<br>
IPDC confirm that this service was activated for the Manila ISP in
December of 2019, right around the time Cloud Innovation's service
with SEACOM, in Cape Town, ended.<br>
<br>
As it currently stands, the ISP in Manila is now off the Internet,
with some 12 paying customers currently without service. Their
excuse - they do not have an AS number of their own.<br>
<br>
IPDC tried to find out why the ISP in Manila thought that it was
okay to use another network's AS number for their service, and as
it turns out, the network engineer at the Manila ISP that set this
up has since left the company. All the ones currently there do not
have any history about all of this.<br>
<br>
Currently, 156.241.3.0/24 is not in the global BGP table:<br>
<br>
*****<br>
<br>
lg-01-ams.nl>sh ip bgp 156.241.3.0/24<br>
% Network not in table<br>
lg-01-ams.nl><br>
<br>
lg-01-nbo.ke>sh ip bgp 156.241.3.0/24<br>
% Network not in table<br>
lg-01-nbo.ke><br>
<br>
lg-01-cpt.za>sh ip bgp 156.241.3.0/24<br>
% Network not in table<br>
lg-01-cpt.za><br>
<br>
*****<br>
<br>
That Cloud Innovation thought it was okay for them to use
MacroLan's AS number to originate their own prefixes into the
global BGP is as morally reprehensible as it is technologically
distasteful.<br>
<br>
SEACOM have been working very closely with AFRINIC to delete
previous route objects from their IRR that certify a relationship
between Cloud Innovation's parent /16 aggregates that cover this
prefix, and AS37353, but this is one of those objects that cannot
be removed via the MyAFRINIC portal, and requires manual
intervention from AFRINIC.<br>
<br>
I have not, personally, spoken to the proprietors of Cloud
Innovation and/or Outside Heaven, as I don't see how anything
could explain this with any degree of justification.<br>
<br>
For now, I will find some beer to wipe the foul taste from my
mouth, while we (SEACOM) consider what to do about this.<br>
<br>
And yes, for those who are wondering, RPKI's ROV would not have
prevented this, in its current form. This is AS hijacking, and
ROV, today, tries to solve the prefix-hijacking problem, first.<br>
<br>
Thank you for your attention.<br>
<br>
Mark.<br>
</font>
</body>
</html>