Hacker News Reader

Comment by caymanjim | Hacker News Reader

I'm one of those 82%. The risk is vastly overblown. If you use apps (for e.g. banking), they're already communicating over encrypted connections, and they're not going to accept unknown certificates. They're completely secure. The Google Play Store and the Apple App Store are completely secure. Your browser is going to warn you about certificates that don't match expectations. You have to go out of your way to fall victim to any MITM attack. The risk from a hostile WiFi is pretty close to zero unless you're the kind of person who blindly clicks on links in email you get, and people like that are unsafe in any environment.

View on HN

> They're completely secure.

No. They are an insecure transport, like most of the internet. The majority of high profile sites, services, and apps will now use TLS by default, but this is one of the few edge cases where a reputable VPN provider would help protect end users from poorly secured apps and services. I would trust the VPN folks vetted by Mozilla, but I am biased because I worked at Mozilla for years and have a great deal of trust for the folks there, or for the technical audience at HN I would recommend using one of the cloud service provider templates for standing up your own VPN server.

How does using a VPN protect you from poorly secured apps and services?

My understanding is a MITM attack can happen between your VPN provider and the service, just like it may work on the wifi point.

You should do your initial vpn setup and first connection over a known good secure connection. Good vpn apps also pin certs and will warn you about migrating hosts or unexpected IPs.

It doesn't, it simply protects the transport between the untrusted network location (e.g. coffee shop, airport, or Apartment 23C's unsecured wifi) and your VPN endpoint, which presumably has a suitably "safe" transport.

Without digging into too much detail, or providing a whack of data because I am not writing long form about this and I spent too much time writing comments on HN this morning :P, but my observation has been that the majority of unsecured apps have service endpoints that are deployed in cloud service provider environments, so if you have a VPN end point that terminates inside the same region, then it is highly unlikely that your traffic will hit the Internet before arriving at it's destination. That said, it's TCP/IP, so all it takes is a routing change and all of those assumptions go out the window...

Beyond that, MITM and traffic monitoring is happening all of the time in many network service providers, not all of it's nefarious -- alot of traffic analysis is just sampling to make sure that the network is working well and to understand whats happening. That said, there are many service providers and jurisdictions that opportunistically capture data or modify traffic in flight for purposes other than quality of service (e.g. ISPs injecting advertisements into HTTP traffic, surveillance, etc)

Using a VPN in this way doesn't protect you against those kinds of threats - the value in using a VPN here is about preventing local eavesdroppers from monitoring network metadata that can be used to classify your traffic and track you.

There are many examples, but one clear one is for marginalized people, such as folks from the LGBT community using specific apps (e.g. gay dating apps such Grindr or Growlr) that could out them to a monitoring service while in a jurisdiction or community hostile to them. Having that traffic tunneled safely to a "known safe" destination, e.g. an AWS or GCP instance in a safer region, would provide a measure of protection.

I had a discussion with a colleague about this about a year ago, but I think it would be very awesome to build an open source VPN service that leveraged AWS & GCP VPCs and inter region peering so that it would egress in the same cloud provider regions as destination IP blocks to prevent as much traffic as possible from flowing over the public internet, but I haven't had the time to put into it.

Would you mind naming a "mainstream" application that does not use TLS by default?

Yes I would mind, but only because I stopped maintaining a list when I stopped working for a browser vendor (Mozilla) and an organization that has an intercepting proxy as a part of it's core product(OpenDNS).

It's not a hard experiment to run, and "mainstream" has different meanings in different locations. Get a femto cell or other remote access point set up an open wireless network in a high traffic area, especially in a place where there is a transit hub, and hang out for several hours.

You could also grab the Alexa top 1 million or Cisco Umbrella top 1 million, and scan them for http endpoints, and work backward from there for sites that don't bounce you to a TLS endpoint.

> I would trust the VPN folks vetted by Mozilla

Does mozilla have a list of "vetted" VPNs? I'm interested in seeing that list.

AFAIK just ProtonVPN right now, but I haven't worked at Mozilla in almost 5 years. Tweet @MozillaSecurity and they will probably point you in the right direction, most of them are super helpful!

My brother uses match.com and he kept getting his credentials stolen. He claimed he had a really good password. It happened every time he made a new account. Came down to using public Wifi. The way match.com did authentication was horrible and some info was in clear text. I didn't believe it, but did a web search at the time and found there were a lot of people in the same boat.

Is it really each app that verifies its own certificate? Or is it relying on an iOS/Android service to verify the chain of trust? I don't know what the UI would look like, but at the time you connect to WiFi couldn't the user unwittingly accept a pushed certificate, functionally similar to the browser add exception button for an invalid certificate warning, but system wide?

I disagree with the comment "go out of your way to fall victim to any MITM attack". If you're trusting employer provided or controlled equipment and haven't personally verified the certificate lists, they totally could MITM your connections without your knowledge if they've already added the exceptions. I know some employers do this, and I know network equipment vendor sales staff have used the ability to MITM employees as an explicit product feature and sales tactic including boasting about the ease with which they can procure faked certs for Google, Microsoft, et al, for this purpose.

The risk to individuals is vastly overblown, but if 82% of people connect to any free WiFi they find and 1‰ of those are passing credentials insecurely, an attacker needs to sample on average 1220 people. That's about a hit a day at a busy coffee shop.

So y'all and I don't see a risk, but a malicious actor sees a valid strategy.

Someone's location also impacts their risk, but generally as long as one isn't in a very dense, high-value-of-information area, the realistic risk that cybercriminals are staked out somewhere like a Starbucks in Topeka, KS waiting to lure people into a MITM is probably less than the actual device getting stolen.

Things have improved lately, but it wasn’t that long ago that the Chase banking app on iOS did not check the certificate hostname at all, meaning that MITM for the TLS it used was trivial.

There are risks, and fortunately they are reducing.

A VPN is still a good idea.

>They're completely secure.

No, they are not. They will never be. They might be secure, but not completely. It is true that we are getting safer, but you should not stop thinking about the risks that will never go away.

I don't worry about being randomly stabbed, that seems to be more likely than someone hacking my https connections on a public wifi.

Just like there certainly are some cities or places you should actually worry about being randomly stabbed, there are some public open access networks that should raise your concerns.

I don't say that you should always be overly concerned when using a wifi. Though, you should pay attention at which wifi you log on and what services you access while doing that.

Do you really need to check you email on a public wifi? If it's not necessary, then just don't do it. It's the small things. Sure you can take this into the extreme (which is never a bad thing in this case), but even if we don't, then we should not teach people that every public wifi is perfectly fine. Please remain thoughtful.

I believe you are very much mistaken. The risk/reward profile of both acts are wildly different: the chances of getting caught while operating a honeypot AP in a public environment are vanishingly small, not so with running around brandishing a knife. Think of it like gas station card skimmers (where the risk is substantially higher than the rogue AP) where the financial motivation is already enough for some people to try it. I might be a tad cynical on this, but since there's money to be made I bet it happens more often than we hear about.

Your threat model should be localized. There are plenty of places where the threat of stabbing is low, and should not occupy your mind. There are plenty of places where the threat of stabbing is high, and you should be aware of your surroundings.

Likewise, your threat of being preyed upon on an open AP is dependent on what you're doing on that AP. Surfing HN? Who cares. But if I'm checking my bank account balance or touching sensitive info, I'm absolutely going to fire up a vpn first.

    > No, they are not. They will never be. They might be secure, but not completely...

You are right in that nothing is ever "completely" secure.

But we take risks ALL THE TIME. In a primitive sense, risk assessment can be thought of as a product...

    risk = severity_of_outcome x probability_of_occurence

Just how risky is it to get on a public wifi? What could happen? What does happen? Are there "safe" things you can do on a public wifi? Are we talking about giving up treasure to casual script-kiddies? Or is it only nation-state-actors that we need to worry about, or maybe jealous-obsessive hacker boyfriends? What can happen if I just leave wifi enabled on my phone? What can happen if I connect and do nothing? What are the exploits?

A lot of these things are unclear-- even for "experts" on HN.

In the absence of actual credible stories about the bad things that happen on public wifi, people will DISMISS the risk. Some may choose a more paranoid approach and never use public wifi, but I think that's just about as unreasonable as the opposite.

You typo an address in your URL bar and an attacker can serve you a fake login page of your intended destination complete with valid SSL cert.

Regular users should not be using the URL bar, they should type whatever they want in the search bar.

And click on the first link, which is a Google ad?

I disagree strongly here. We should be encouraging users to take responsibility and agency for their actions. As admins and tech-aware people in our circles, we should be constantly encouraging better habits like typing URIs manually, decreasing the use of search engines for non-search purposes, and trying to spread literacy.

Besides, it's not like we should ever trust a search engine as an authority on anything but the results that bring the host company more money.

Why is typing a URI manually better?

It's undoubtedly less secure and it's upside is...what exactly?

Personal responsibility is the main benefit; teaching people to look twice at what they do on the Internet. And frankly speaking, if manual user entry is less secure than software that the user does not own from metal to UI, then we need to upgrade the users, not the Web. Demanding a manual entry means that the user is taking deliberate action, not following whatever some piece of software is telling them to be correct.

Personal responsibility, in and of itself, is not usually considered to be a benefit. It is considered a thing one has to undertake to achieve a benefit.

You seem to be suggesting that if I intend to visit mywebsite.com, that typing "myw", seeing "mywebsite.com", and hitting enter is somehow less deliberate than typing "mywebsit.com" and going to a site that was not my intent.

Given that only the first one of these things reflects my intent, describing the second as deliberate and the first as not-deliberate requires some odd twisting of definitions.

Your fear, it seems to be, is that tools we use might influence how we act. This is nothing new. Stories started to rhyme less when we figured out how to write things instead of memorizing them. That was still probably undoubtedly an improvement. So can you perhaps clarify what specific influences that our tools have might be bad? For example, my browser suggesting "mywatertower.com" instead of "mywebsite.com" because the first paid for a higher position.

That seems a reasonable end state to fear, but I have no reason to believe we're heading that direction. Do you? Is inconveniencing (literally) billions of people and forcing them to take less secure paths to do what they want worth avoiding a possibility that certainly doesn't seem imminent?

To me, it's definitely worth it to "inconvenience" people and demand conscious behavior. With the largest browser vendor being the largest advertising agency and the largest search engine, encouraging anyone to have anything beyond the barest minimum to do with them goes against everything I believe in. Search engines should not be able to intercede in direct Web activity; they should be used for content discovery and nothing else. You should not be able to type three letters and get a full URL unless you personally wrote a macro to do that.

Then again, when users stopped having to specify protocol for every server, that was probably the beginning of the end of deliberate browsing. My fear is really that people take the Web and the Internet as a whole for granted; and I want to see procedures put in place to demand that all users be aware of their actions, what data they share with servers, and all of what's being downloaded to their computer.

I don't like having to remind people that there is no "Cloud", only someone else's computer. And you're saying you trust that other person's machine and security more than your own by storing critical data there. Or pointing out what having one CDN for so much of the Web did last month.

> You should not be able to type three letters and get a full URL unless you personally wrote a macro to do that.

Should you be unable to drive unless you can, IDK, build an automatic transmission? This point of view reduces, as far as I can tell, to the idea that good user interfaces should only be extended to those privileged with enough expertise to build them themselves.

This requires a level of literacy that most users will never be able to meet, by virtue of most people not having time to learn a programming language since they have other responsibilities.

It's also not clear why you aren't extending this backwards: why is it okay to use a browser I didn't write myself (or at least compile from source)? What about my OS? Do you really expect every user to have full knowledge of their entire system? That puts severe limits on the potential tooling we can use, and goes counter to one of the core ideas of software engineering: abstraction.

> And you're saying you trust that other person's machine and security more than your own by storing critical data there.

Yes, I trust security teams and engineers whose job is security and reliability more so than I trust myself, in much the same way that I'd trust a surgeon to do surgery better than myself. I realize that there's certification/training differences that may be relevant, but in general, the same ideas apply.

Manually typed URIs lack spell check. Google has this built in. The probability of Google not taking you to Chase but some malicious site is far far lower than a typo of a manual url being registered.

That's already the case whatever local network you happen to be using.

I connect all my usual sites with history completion from the address bar. I never type them.

I don’t know if you are right, it depends on whether your word “typo” is a typo - itself a very meta thing!

“You typo an address” — well, then, sure. But how did you manage to typo an address that’s exactly where the attacker put their stuff, and the bank didn’t catch them?

Now, if you meant “type”, then NO. Unless the browser includes a certificate authority that gave out a certificate to the attacker for that domain, the browser will flag this “valid” SSL cert.

No, attacker can just MITM your DNS traffic and synthesise lots of typosquats of their own. They can't do the actual bigbank.com or facebook.com domains thanks to various protections like HSTS, but typos are unlikely to have HSTS preload etc, and they can have vast numbers of them at basically zero cost.

Solved by using something like cloudflare's 1.1.1.1?

Not at all, since the DNS traffic to 1.1.1.1 is still unencrypted and unauthenticated (unless you're using DNSCrypt or similar, or the domain you're looking up uses DNSSEC and you're using a validating resolver).

Correct me if I'm wrong, but I think the ios app has dns over https or dns over tls enabled.

No idea; I took your comment to refer to using 1.1.1.1 for DNS in general, since it didn't mention any app.

But will my credential autofill put in my creds? I think people are still safe.

There is always that amateur app you thought you deleted that uses http to send your credentials in plaintext.

I don't really care if my chess app is insecure.

But now my opponents will know I only ever open with d4.

You will when the hackers checkmate you.

You probably use a different password for every login, but there are a lot of people who don't.

Reply on news.ycombinator.com