Show HN: Responding to NSA spying with a simple consumer VPN service | Hacker News Reader

Show HN: Responding to NSA spying with a simple consumer VPN service

65 pointsrdl74 comments

The NSA metadata gathering and passive monitoring is quite upsetting, and we don't have much confidence in legal changes happening quickly, so we put together a really simple consumer VPN service. We're still cleaning up the UI and making the install process easier, but it should work for people now.

Yes, it's based in the US, but there's a big difference between "will turn over data proactively" and "will push back on requests to the fullest extent of the law". Since we don't fall under CALEA, there's no requirement for us to have any monitoring infrastructure.

We're focusing on the mobile experience for iOS and Android -- the best combination of platform security but also difficult to "roll your own" service.

Would greatly appreciate HN's feedback on concept and implementation; still under active development. Posting some free signup codes in the comments to try it out.

https://privacy.cryptoseal.com/

View on HN

The Biggest problem I have found with VPN providers is bandwidth. I just installed DD-WRT on my router to use OpenVPN client on the router, the bandwidth speed drops to a level that makes the setup totally useless.

It might be because of consumer grade router[1] have low computing power, but the speed is so low that I highly doubt it. Yes I have tried multiple VPN providers.

I have yet to find a VPN provider that can provide all of the following:

- Guarantee Bandwidth

- Take Privacy Seriously[2]

- Support consumer grade router (with DD-WRT or alternative)

- Do not cost more than my actual internet connection

[1] ASUS RT-N16

[2] http://torrentfreak.com/vpn-services-that-take-your-anonymit...

What kind of performance are you looking for, and what traffic mix?

Testing various consumer routers with VPN for performance is on my todo list. I know the WRT54GL is probably fairly antiquated at this point, but I have some newer DD-WRT supported routers like the WNDR3700 to try.

I am talking about normal every day usage Browsing, Youtube, Netflix, etc from multiple devices at home.

Maybe it is hardware as mention in other replies. I haven't done testing with different hardware.

> I just installed DD-WRT on my router to use OpenVPN client on the router, the bandwidth speed drops to a level that makes the setup totally useless.

A fair amount of that has to do with how little CPU is available on your router device, I suspect no manufacturer will support a device that can do a lot of crypto and still be in the consumer price range.

That said, I could ship you a device right now that'll do 100Mbps of encrypted traffic all day long, but it doesn't have wifi and its about $500...

For "roll your own" with an embedded device, the Marvell Kirkwood is probably the best since it has AES acceleration in hardware. One of the Sheevaplugs has this. It's not really well supported in Linux yet. It ends up being about as fast as an Atom.

The Kirkwood AES accelerator (mv_cesa) is not exposed to userspace in Linux. It won't accelerate your SSL. You can use it to accelerate dm-crypt, though. I owned a Sheevaplug and own a Dreamplug. They're not nearly as fast as even a 2009 Atom... But power-efficient, yes.

Is that just a driver limitation, or is there a problem with the hardware so the driver is written that way? It also doesn't do DMA, so I've heard bad things about performance with it for small packets.

If I remember correctly it's a driver limitation, and theoretically it could be modified so it would do what you want, but consider the amount of userspace/kernelspace crossings to be made just to send some ciphertext over the network to some other party.

I imagine something like this would take place:

SSL does not happen in the kernel. So you need to send a payload to the kernel to encrypt, you get back the ciphertext, wrap it in your packet (add some headers) and then push this data once more to the kernel, which sends it out over the network for you.

Receiving ciphertext would go along similar lines.

That's probably not very efficient. Contrast this with full disk encryption using dm_crypt (linux). No shuffling back and forth of data. Userspace only really needs to handle the plaintext (and not even directly, but through the FS layer, which is also in the kernel).

So maybe it may just not be worthwile to expose the interface to userspace on reasonably powerful platforms? Anything above a tiny MIPS or old 486?

The VIA padlock crypto accelerator is (or was at some point) exposed to userspace and OpenSSL could take advantage of it. I wonder how much of a performance increase that yielded, considering the issue I noted above.

Also, hardware accelerators typically only support a couple of modes and key sizes. That's fine for full disk encryption, but in a networked world you need more flexibility.

> I wonder how much of a performance increase that yielded

It yielded 100Mbps+ SCP's with zero system CPU usage, very very handy. I've used a few older SSL accelerators and they were good for about 1Gbps or so. IMHO AES-NI makes these things obsolete, and with that going into the new Intel Atom cores, I think thats likely going to drop a lot of the competition out.

Ah, yes, the Silvermont series should be amazing. It looks like it might have virtualization extensions, too. I can't tell when it's supposed to come out, though.

Should rock for NASes too. Right now only certain i3/i5 NAS seem to do AES acceleration, so the cheapest are around $2k bare from QNAP or $999 from iXSystems.

You are correct. I use OpenVPN to terminate a connection and get great performance, but it's served up by a reasonably powered box. If OP is using dd-wrt, I'm being it's on a WiFi router with 300mhz cpu at best and 48-64 mb ram.

I'm sure there are many legit VPN provider out of there, which really care about the user and provide some hardware/bandwidth, but most are oversold services (barely scam) with very high price for what they offer. I think the solution is a self-made VPN out of a VPS. VPS nowadays are really cheap, and you can get awesome deals with some very serious providers. Also, you can easily get one outside the USA, for (in my opinion at least) increased security.

For anyone interessed, https://github.com/Nyr/openvpn-install makes installing OpenVPN a breeze. For VPS deals, http://www.lowendbox.com/ and http://vpsboard.com/.

Absolutely agree. For those who have trouble setting up OpenVPN (it's a pain in the ass) and want a simple solution, sshuttle works really, really well:

https://github.com/apenwarr/sshuttle

Running your own VPN server means no privacy, as you are only user of the system and all the traffic going in/out of the server is attributed to you.

What VPN providers claims is that they do not store logs/ip addresses and they have no way of knowing what traffic comes from client.

Did you try IVPN? http://www.ivpn.net/

I've just purchased an Asus RT-N66U to run an OpenVPN client so that I can encrypt the traffic from all computers, and as I don't do file-sharing I've focused on choosing a provider obsessed by privacy rather than piracy. IVPN looks like it, but I haven't yet tried it so don't yet know how it will perform.

Has anyone set up a Raspberry Pi as a router?

Been there, low performance :(

You might look at https://torrentfreak.com/which-vpn-providers-really-take-ano...

I've been using one of these, bought with Bitcoin, via Coinbase, set up with a Hushmail account, for a while now. It works great. Bandwidth is diminished somewhat, but not enough to be a problem for normal browsing or other totally legal activities.

edit: I'm tired, just noticed you linked to essentially the same TF article I did. Anyway, the service I've been using works with Tomato and DD-WRT (on some routers), is affordable, and is on of the recommended services for taking privacy seriously. They don't guarantee bandwidth, but in practice, there hasn't been a problem so far.

For the tech details: it's OpenVPN, PPTP, or L2TP, your choice. Yes, PPTP and L2TP have "issues", but they're currently the easiest way to get a VPN on iOS without a custom client.

These "issues" are critical in thwarting PRISM. More security snake oil.

Well, we don't control iOS, so to support better quality VPN on iOS, you need to have a custom client, which we don't yet have. I was optimistic iOS 7 would have some SSL VPN standard built in, but it doesn't.

You can use OpenVPN, which is pretty robust (SSL-based) on any platform where it's supported. We're working on an SSL VPN option for mobile.

The PPTP and L2TP vulnerabilities aren't purely passive attacks, so they're not likely to be used by NSA for general users. They are a risk for individual targeted users.

If enough people make it difficult enough to beat those implementation "issues" then you still have security in the aggregate.

The Soviet soldier was almost always outclassed on a 1:1 basis by the German soldier in WWII, but the Wehrmacht and Luftwaffe combined still were unable to defeat the U.S.S.R.

There's an official OpenVPN client on iOS now in the App Store that will accept a standard .ovpn profile.

Wow -- can't believe we overlooked that. Adding, and will probably make that the default.

I'm not sure how this is at all relevant to the NSA spying - the NSA are gathering details from content providers (like Facebook, Google, Skype), so if you use the VPN to access any of them you are still at risk, or if you communicate using with anyone using one of those services.

It's not the whole solution, but the vast majority of NSA spying is passive "over the wire" stuff. We know they do that, pervasively, and even more, we know foreign countries do that (and filtering, and in some cases lots of fun tampering) on connections (e.g. Great Firewall of China). NSA has been doing passive intercept since inception in the 1940s.

PRISM is somewhat ambiguous -- maybe it's a huge secret program where they get blanket access to big sites directly, maybe it's just a UI layer for managing subpoena or warrant results.

Protecting your metadata is one thing where a VPN works pretty well. There are still some more advanced attacks (looking at the encrypted traffic flows on lightly loaded links, you can infer what site/activity one is doing, even without decrypting, unless you pad all communications).

If the NSA is actually doing widespread monitoring/collection of backbone traffic will a VPN actually help at all? End-to-end encryption seems like right solution.

EDIT: nevermind, VPNs do offer obfuscation of "metadata", i.e. your IP address, which I guess is what you were saying.

I wonder if you could create a VPN/Tor-like network that automatically picks "exit nodes" closest to the destination to avoid taps... (or even better, avoiding known tapped routes)

We actually do some pretty interesting internal routing for that, but only have 4 geo locations so far. Adding more, and adding some "unique" security features in the future, but wanted to get something out there now. Doing "cross border" that way is probably the most meaningful, assuming many countries set up their monitoring on international links. Being aggregated inside a VPN service helps a lot there.

End to end is mostly the right solution, but I'd like to see message/object based encryption (conf and signing) rather than just transport encryption, too.

(other metadata includes DNS resolution, and of course all the other protocols and things which aren't encrypted -- it's not JUST an issue of hiding your IP)

So it's possible different hosts will see different IPs for me depending on where they are in the network relative to your locations?

Yep! Ideally you don't want to use different IPs while visiting the same site because they'll likely block your account, but for sites on different sides of the globe, this is going to happen.

> I wonder if you could create a VPN/Tor-like network that automatically picks "exit nodes"

Yes, we do that. We have some internal traffic balancing and mixing, we will probably put more detail to that before the end of the year.

We are looking at the options for VPN-to-Tor as well, however there are some performance challenges to overcome.

It's not clear that this is what's happening. Especially since the companies have vehemently denied such allegations.

One thing the NSA fiasco has done is made people aware of the security and privacy issues on the internet and it's a great time to launch a service like this. Hopefully we will have most of the internet users using VPN (and hopefully Tor).

I also love the fact that they have servers outside of the US, makes it a little bit harder for the US government to spy on you.

Yes, we love Tor, and are looking at ways to help Tor and make Tor more robust and easier to use as well.

It would be ironic if the NSA fiasco ended up accomplishing half of the NSA's mission (protecting domestic networks) by getting everyone to improve security and encrypt-by-default, at the cost of making the NSA's SIGINT mission vastly more difficult.

Coupon codes: hn50hnaeS6SaeR is 50% off for HN.

What's the difference between you and eg. http://unblockvpn.com/ who offer easy VPN access for less than $5 a month?

Or how is your product easier to use?

There are ~hundreds (thousands?) of VPN providers worldwide. I've seen as low as $1-3/mo -- and some which are ad sponsored and free. I haven't used them all, so I'm not sure of the specific advantages and disadvantages of each.

We have some differentiating security features vs. all existing VPN systems coming; this is just a show hn to get feedback.

We've mainly worked on the corp VPN as a service (which does internal monitoring/filtering/IDS/etc.), but at $100/mo, it's not really an option for consumers. Due to all the NSA stuff, people were asking for a consumer option.

I intend absolutely no offense here. I love seeing HN projects posted here. But...

One of the major issues that has been raised recently is, essentially, trust. Especially in the market that you'd be targeting -- individuals that no longer trust various online services.

You can say that you will fight requests as much as possible under law, but how is that different from what Google, Facebook, and others claim to do?

The NYTimes just published an article claiming that Skype was backdoored by the NSA in cooperation with a small team of Skype developers, in secret, back in 2011. For the sort of people that are concerned by that sort of news, how are you going to convince them that you're different?

This exchange may provide some insight... https://news.ycombinator.com/item?id=5914546

Thanks. I was also alluding to the service being compromised in the future though. i.e., whether or not they've come up with something new, above and beyond the usual assurances.

Yeah, we're actually working on a couple of things which will largely solve the trust problem (which is a bigger thing than just VPN service); stay tuned.

For now I'd like to think we're more trustworthy than a large/general purpose company because we have a lot less to lose in fighting (and much more to gain).

One of the items I'm going to be most interested in from your service (when the time comes for you guys to start working on it) will be the TOS and/or practices you put in place.

I've been a ViprVPN customer before. I had a question or perhaps it was an issue I called them about and the person I was communicating with told me what VPN server I last connected to and when I connected. Sure, to do any kind of troubleshooting, this would have been necessary and important information. But I was concerned enough about the unsolicited disclosure that I cancelled the service immediately.

DuckDuckGo can claim a reasonably high interest in protecting my privacy because they simply do not collect data that the big search engine does. Collecting and storing this data would make them a target for undisclosed, unchallengeable, and unwarranted surveillance. This has enormous appeal to me.

Having said that, have you guys discussed (loosely) what data you will be collecting?

Our goal for the privacy service is to collect as little as possible. (the corp service is totally separate tech and infrastructure and has user-configurable logging). We're trying to figure out what the absolute minimum is. We're also looking at Bitcoin and other forms of payment.

For a $5-10/mo VPN, we're probably going to handle most problems by "open a new account, here's a service credit", so we don't actually need to debug much. We have a vested interest in collecting the minimum information possible so there's no point in subpoenaing it from us.

Thanks for making this viable for noobs like me.

There are already plenty of equally viable options for cheaper. I hear good things about https://www.privateinternetaccess.com/pages/client-support/ , and it is about 1/3rd the price.

Howdy, CryptoSeal guys. It was great grabbing a beer with you in SFO earlier this year and talking about the VPN space. GetCloak.com continues to go well. I didn't realize you were heading in the consumer direction. We should probably catch up again someday soon! ;-)

I like the idea of more VPN providers coming online. One thing that may be helpful is to differentiate yourselves from the other VPN providers. For example, I have a VPN provider. Should I switch to cryptoseal.com?

Honestly if you're happy with your VPN provider right now, it's probably not worth switching yet. We try to do the baseline VPN service as well as possible (performance, support, etc.), but it's not that different from other providers, so please consider us if you're not happy with the current provider, but otherwise you're probably just as well served with whatever you're using now, for now.

We're working on some things which will make it compelling to switch. We put it up now because a lot of people don't have VPNs today -- so hopefully adding another provider convinces some additional people they could use a VPN.

Without having visited your URL, or read your TOS, or evaluated your service for ease of use / viability, I will tell you right now that I would gladly pay a cost equivalent to what I pay monthly for broadband for a VPN service that's reliable and can protect 100% of my Internet traffic in transit. (Exit point, destination points are a whole 'nother animal.)

So, since you sound like a somewhat higher end user:

1) What platforms do you care about? Do you mainly need service from one fixed location, or from home/office network plus mobile?

2) How close does it need to get to the endpoints? We have 4 exit nodes right now; we'd probably need ~50+ to be very close to most services. There's still a portion which is "in the clear" (although, use https...), but it becomes very impractical for NSA or especially others to passively tap all those locations (since they wouldn't be IXes necessarily, and intra-colo links don't get routed through buildings like ATT 611 Folsom St.

Actually, with sufficient BGP peering, you can probably reach most major destinations with far less than 50 (CDNs do), and you increase the likelihood that traffic never traverses a telco backbone.

There are 3 platforms I care about, Windows, Android, and Linux.

For the last 2 weeks I've been taking actions that attempt to pull-back my public footprint and re-exert control over my privacy (admittedly illusory) to a point where I feel more comfortable. One of the biggest "oh crap" moments was when I realized my phone is powered by software derived from a Google product.

Windows desktop should be straight forward enough to connect.

Personal Linux server farm with services that are open to the general Internet, but for outbound traffic not related to something I serve, my servers are at the mercy of the security of the feed I have.

As far as how close it has to be to an exit, I'm completely indifferent. There are trade-offs that I'm willing to accept (in some cases, extreme) as I record over 20 years of habitual internet behavior.

If you have a Galaxy Nexus S or similar you could try the FirefoxOS image, the base Linux system on there is quite simple and everything seems to be available in Github. Of course you still have the baseband to worry about, but apart from a few OMAP850s you're going to have a hard time getting away from that.

Do you do ANY kind of logging?

We do logging on our web server, obviously the support ticket system, our mail server, coupon code redemption, etc. Billing with Stripe -- presumably they keep records for a long time. We don't see your CC, but are notified of payment, etc. (which is why Bitcoin is very attractive to add, and maybe other systems.)

The privacy VPN itself currently does zero logging. The best practices seem to be either zero logging or very short retention logging. We'll commit to one of those (but most likely zero logging) soon (working on a very clear and plain language ToS). All the stuff we'd handle with logging is instead done by going out to top-500 sites (or anything reported to us as not working), rather than monitoring use.

We don't currently do "anonymization" so web browsing can be an issue. We're looking at that with some kind of opt-in proxy.

> We don't currently do "anonymization" so web browsing can be an issue.

What does that mean exactly?

Browsers leak information. Profusely.

We have looked at various ways to solve this, but generally the easiest thing breaks a bunch of other "stuff". We'd like to not break stuff, but still be private. We have a couple strategies but haven't deployed them yet.

We are talking about a VPN here aren't we? Won't anything the browser "leaks" go down the same pipe and get spat out your end? Assuming your using In Cognito and connecting to TLS websites...I can't imagine how there could still be a significant privacy risk.

He means that the browser leaks information to the website that it is connecting to. More info:

https://panopticlick.eff.org/browser-uniqueness.pdf

Payment is by Stripe, how long are records of the transaction kept by yourselves? Are you associating a card transaction to an IP address in any way that could result in you ever having to release that data?

We're adding other, more anonymous payment options, which is probably the best solution.

Where are the VPN servers located geographically? It might be wise to include where your servers are located - and perhaps even a test IP address that users can use for latency / packet loss testing.

US-West, US-Central, US-East, UK right now. Adding some more in Europe and Asia ASAP.

A (current and historical) performance monitor, uptime monitor is a great idea.

When will the Sealand node come online?

Thats also a good question, we'll look into it.

Since it's ethernet-distance to the UK node, ... (after I left, HavenCo ran servers out of London Telehouse for a long time claiming they were "on Sealand". It was lulz when the place caught on fire and everything was unaffected. Previously, we had 4xE1 running from border routers at Telehouse back to shore near Sealand and then a microwave link, but I guess they figured this was cheaper and more reliable and no one would check... :)

Good news, I actually put latency/loss monitoring into our internal support system last night, so I'll probably be exposing some data from it next week.

We currently monitor 60+ sites from the Alexa Top 500 for load time and loss so we can use the data for capacity planning and fault isolation.

you broke #1 rule of the internet: "Logo in header must point to the homepage."

At least the text next to it does, but will fix -- the traditional error is "blog header links back to blog vs. corp site" which we've solved for now (by not having a blog).

For the noobs, what's the difference in OpenVPN between UDP and TCP?

UDP will give you better performance. TCP will work better in pathologically bad network conditions (either some kind of firewall, or some specific kinds of network loss).

The issue is that "TCP in TCP" can lead to weird interactions where you delay one packet, wait for retransmit, etc., and essentially a single packet lost can eat up a second or two.

In general I'd always try UDP, and if it doesn't work, fall back to TCP.

They're equivalent security -- it's just network performance.

Thanks! Good to know. Service worked flawlessly.

I assume you know what a handshake is, right? A TCP connection requires a handshake and is evidently a much more reliable and safe connection. A UDP connection doesn't require a handshake. This means that you never know if your data has reached its destination or not and due to this is much less reliable. Skype uses a UDP connection, and at times it is evident that it does when video resolution drops, etc.

How about one of those tryout codes?

Reply on news.ycombinator.com