I started drafting this post just before I left Ireland for Enigma 2017. While at ENIGMA I realized how important it is to write about this because it is too damn easy to forget about it altogether.
How secure and reliable are our personal infrastructure services, such as our ISPs? My educated guess is, not much.
The start of this story I already talked about: my card got cloned and I had to get it replaced. Among the various services that I needed it replaced in, there were providers in both Italy and Ireland: Wind and Vodafone in Italy, 3 IE in Ireland. As to why I had to use an Irish credit card in Italy, it is because SEPA Direct Debit does not actually work, so my Italian services cannot debit my Irish account directly, as I would like, but they can charge (nearly) any VISA or MasterCard credit card.
Changing the card on Wind Italy was trivial, except that when (three weeks later) I went to restore to the original Tesco card, Chrome 56 reported the site as Not Secure because the login page is served on a non-secure connection by default (which means it can be hijacked by a MITM attack). I bookmarked the HTTPS copy (which load non-encrypted resources, which makes it still unsafe) and will keep using that for the near future.
Vodafone Italy proved more interesting in many ways. The main problem is that I could not actually set up the payment with the temporary card I intended to use (Ulster Bank Gold), the website would just error out on me providing a backend error message — after annoying Vodafone Italy over Twitter, I found out that the problem is in the BIN of the credit card, as the Tesco Bank one is whitelisted in their backend, but the Ulster Bank is not. But that is not all; all the pages of the “Do it yourself” have mixed-content requests, making it not completely secure. But this is not completely uncommon.
What was uncommon and scary was that while I was trying to force them into accepting the card I got to the point where Chrome would not auto-fill the form because not secure. Uh? Turned out that, unlike news outlets, Vodafone decided that their website with payment information, invoices, and call details does not need to be hardened against MITM, and instead allows stripping HTTPS just fine: non-secure cookies and all.
In particular what happened was that the left-side navigation link to “Payment methods” used an explicit http:// link, and the further “Edit payment method” link is a relative link… so it would bring up the form in a non-encrypted page. I brought it up on Twitter (together with the problems with changing the credit card on file), and they appear to have fixed that particular problem.
But almost a month later when I went out to replace the card with the new Tesco replacement card, I managed to find something else with a similar problem: when going through the “flow” to change the way I receive my bill (I wanted the PDF attached), the completion stage redirects me to an HTTP page. And from there, even though the iframes are then loaded over HTTPS, the security is lost.
Of course there are two other problems: the login pane is rendered on HTTP, which means that Chrome 56 and the latest Firefox consider it not secure, and since the downgrade from HTTPS to HTTP does not log me out, it means the cookies are not secure, and that makes it possible for an attacker to steal them with not much difficulty. Particularly as the site does not seem to send any HTTP headers to make the connection safe (Archive.is of Mozilla Observatory).
Okay so these two Italian providers have horrible security, but at least I have to say that they mostly worked fine when I was changing the credit cards — despite the very cryptic error that Vodafone decided to give me because my card was foreign. Let’s now see two other (related) providers: Three Ireland and UK — ironically enough, in-between me having to replace the card and writing this post, Wind Italy has completed the merge with Three Italy.
Both the Threes websites are actually fairly secure, as they have a SAML flow on a separate host for login, and then a separate host again for the account management. Even though they also get a bad grade on Mozilla Observatory.
What is more interesting with these two websites is their reliability, or lack thereof. For now almost a month, the Three Ireland website does not allow me to check my connected payment cards, or change them. Which means the automatic top-up does not work and I have to top-up manually. Whenever I try to get to the “Payment Cards” page, it starts loading and then decides to redirect me back to the homepage of the self-service area. It also appears to be using a way to do redirection that is not compatible with some Chrome policy as there is a complicated warning message on the console when that happens.
Three UK is slightly better but not by much. All of this frustrating experience happened just before I left for my trip to the USA for ENIGMA 2017. As I wrote previously I generally use 3 UK roaming there. To use the roaming I need to enable an add-on (after topping up the prepaid account of course), but the add-ons page kept throwing errors. And the documentation suggested to call the wrong number to enable the add-ons on the phone. They gave me the right one over Twitter, though.
Without going into more examples of failures from phone providers, the question for me would be, why is that all we hear about security and reliability comes from either big companies like Google and Facebook, or startups like Uber and AirBnb, but not from ISPs.
While ISPs stopped being the default provider of email for most people years and years ago, they are still the one conduit we need to connect to the rest of the Internet. And when they screw up, they screw up big. Why is it that they are not driving the reliability efforts?
Another obvious question would be whether the open source movement can actually improve the reliability of ISPs by building more tools for management and accounting, just as they used to be more useful to ISPs by building mail and news servers. Unfortunately, that would require admitting that some times you need to be able to restrict the “freedom” of your users, and that’s not something the open source movement has ever been able to accept.