There’s a growing chorus of voices in infosec these days, calling for the effective deprecation of unencrypted traffic over the web. The only useful purpose port 80 serves these days, they argue, is to redirect to port 443.
Their arguments tend to run along these lines:
- HTTPS is the right way to handle encryption on your website (and other services, too). That’s true. It’s a well-established, well-understood standard that has gone through some awkward moments, but has emerged from the process as a robust and easy-to-use security layer on your web server.
- HTTPS protects you from unethical ISPs/states/actors who like to inject your traffic with unwanted ads and other nasties. That’s somewhat true, but requires a little unpacking, because there are some notable edge cases:
- Many corporations install their own certificate authority on approved devices, and use their own certificates to intervene in communications between you and other secure sites. This allows them to sit in the middle of your communication with your bank, for example.
- Anyone who can get access to either of the end points can simply watch the unencrypted traffic on either end, rendering any security measures in the communications protocol moot.
- Setting up and breaking down a secure session is an extremely difficult process that is well-understood by a very few people. The theory is clear enough, but applying it in practice can get hairy. The 2016 WPAD exploit is a prime example. It was demonstrated that you could use a near-obsolete, insecure method of discovering your network’s proxy settings to eavesdrop on secure communications. (Of course, if you could manipulate that setting, you could pretty much do what you want to the browser anyway. The point here is that there are ways for people to eavesdrop on your secure communications and leave you none the wiser, with or without HTTPS.)
- Establishing the identity of remote sites is still a bit fraught. It’s better today than it was in the past, but an SSL cert on its own is insufficient in the absence of a broader, systematic web of trust. These things require international cooperation, protocols and standards. Not all parties want the same thing, and few indeed are focused only on what’s best for the end user’s privacy. HTTPS Everywhere doesn’t move the markers in this part of the debate.
- Cross-site tracking cookies are a much more pernicious and troubling way for unscrupulous commercial actors to inject themselves into your day to day web traffic.
None of these are arguments against the use of HTTPS. You should use it when you can. Just because the other guy’s going to win doesn’t mean you shouldn’t at least make him work for it. But these are considerations that have to be borne in mind when people start suggesting that a web with HTTPS everywhere is a safer, more private place for everyone.
- HTTPS is growing in use. Get on board! That’s an argumentum ad populum. It’s a useful observation—and a Good Thing—but not reason to move in and of itself.
- Plain old HTTP is insecure, and untrustworthy. The first part is true, and people should know it. Google Chrome’s recent decision to mark all plain text websites as insecure is a good move. But the question of trust is more a practical than a technical issue. In other words, you can communicate securely with an untrustworthy information source, and you can communicate insecurely with a trustworthy source. Trust between end points needs to be distinguished from trust in the communications medium. Technical experts who advocate for HTTPS everywhere sometimes confuse the two. Non-technical people often do. Which means we have to be extremely circumspect about what we promise when we talk about making things more secure for them.
- HTTPS is secure and will protect you from eavesdropping. Yes it is, and no it will not. Yes, the protocol is solid. It’s really easy to do right (which is incredibly important for security), and it works for the purpose intended. But it’s only one part of a much, much bigger picture. As Bruce Schneier is fond of saying, the bad guys only need to find one way to win. Most systems have more than one way in. Many systems allow state-sanctioned eavesdroppers in through the front door.
- Encrypting even normal traffic means that you’re protected from state surveillance across the board. If they can’t differentiate your traffic, it makes it harder for them to single out the things they object to. Sad to say, the real world doesn’t work that way.
If a state wants to watch what you’re accessing over the web, they will find a way. At very least, they will know the end points you’re connecting to. If they’ve singled you out for observation, you’re going to face issues even if you use TOR, VPNs and other tools. Between compromised devices, networks and information providers, your options are quite limited if people decide to watch you individually.
If you’re not being singled out, they probably don’t care what you’re looking at. They might care about individual sources, but in cases like that, they’ll just block the site sources.
But wait—what about someone who’s not yet been targeted, and doesn’t want to be? Good question. There’s a marginal case there, but again, using HTTPS Everywhere is probably not going to be the decisive factor. It sure isn’t in China.
There are very good reasons to support the spread of HTTPS. Anybody who tells you otherwise is just… wrong. You should use it when you’re doing anything that involves your personal information. That includes everything from chatting with friends about what movies are good to transferring millions into your superyacht fund.
But to conclude from this that ‘the only purpose for port 80 these days is to redirect to HTTPS’ is a bit naïve, sad to say. It assumes that there are technical solutions to social/political problems. That’s not necessarily true:
- We’d be vastly better off with regulatory/legislative intervention to stop ISPs and others from messing with your web traffic. The reason for this is that removing the mandate to inject gunk into your web traffic is a far more effective way of circumscribing what ISPs are allowed to do. Let’s call it (part of) Net Neutrality. With or without HTTPS, we still need this regulation. Politically, cynical ISPs can point to HTTPS Everywhere and use it as an argument against net neutrality regulation. It’s not entirely rational, but that’s not to say the argument would be ineffective.
- When changing collective behaviour, it’s often better to proscribe the behaviour rather than prescribe the technical cure. People will always seek ways to fulfil the letter of the law and still achieve their own selfish ends. So an emphasis on better law is generally more effective than emphasising better tech.
- State actors and other surveillance groups generally don’t care that you’re looking at public access websites. When they do care, they just block the site at the ISP level. Yes, there are edge cases where you can get in before they know it matters, but that’s true with and without HTTPS. Encryption does add a slight advantage when it comes to internet whack-a-mole, but it’s only slight. And it’s likely to be ephemeral.
- If someone really wants to track you, they will compromise your end points, not the network layer. This is what China does. They just sit people inside the offices of the main online services. Law enforcement and signals intelligence agencies do much the same thing. They’ll either compromise your devices, or they’ll compromise the other end-point, often with the assistance of the service provider. The FBI attempted this with TOR, with partial success.
- HTTPS is easy, yes, but easy is not the same as simple. Just because you’ve got your cert properly set up doesn’t mean you’re safe. Focusing on HTTPS to the exclusion of other considerations or overselling its benefits could create a false sense of security. It’s far easier to designate something ‘untrusted’ than it is to determine that the same thing is ‘trusted’. We need to be careful about what we’re selling when we say that HTTPS Everywhere makes us more secure.
Widespread use of HTTPS is a Good Thing, and should be encouraged. But mandating its use everywhere is of limited additional utility where your practical security is concerned.
In a nutshell, universal HTTPS alone is insufficient to change or curb malicious human behaviour; and the additional measures that are necessary don’t require HTTPS everywhere to succeed.
HTTPS everywhere would achieve only a marginal and possibly ephemeral gain. In practical terms, people who are most vulnerable to unwanted collective or individual surveillance by state actors gain very little from this.
But let’s keep perspective: Opposing HTTPS Everywhere is a foolish waste of time and effort. We should encourage its spread. What we should NOT do, though, is pretend that the end of port 80 is the end of our privacy concerns, or even a particularly notable win. It’s a marginal improvement at best.
My beef is with people who think this is a big win for privacy. It’s emphatically not.