Hacker Newsnew | past | comments | ask | show | jobs | submit | obnauticus's commentslogin

I originally wanted to do this but the CVE history is a bit too colorful for something I’d want to trust as a “cloud replacement”:

https://cve.mitre.org/cgi-bin/cvekey.cgi?keyword=nextcloud

A common misconception IMO is that running and owning your own infrastructure is somehow more secure. To that I lol, and I’m confident that the thousands of AWS/GCP/Azure/iCloud security engineers are all doing a more thorough job than you can. At the very very least they receive embargoed bugs which they often mitigate before the general public.


One doesn't have to expose it to malicious actors. It is most-useful that way, sure. Mine is at 10.27.0.68. Have fun, hackers!

Also, I lol at most CVEs. Butterfly farted outside, oh uh.

Take the top one: In Nextcloud Desktop Client 3.13.1 through 3.13.3 on Linux, synchronized files (between the server and client) may become world writable or world readable. This is fixed in 3.13.4.

You mean to tell me a few minor point releases imitated umask, making world-readable [and possibly added writable]? Oh no! The tragedy! Keep in mind most clients are single user systems anyway.

Judge them on their facts, there are vulns and then there are vulns. CVEs are a sign of attention on a project. No more or less.


I find that one concerning in an enterprise setup (which they target). Or the fact that the desktop client has 999 open issues. Or that the last version silently takes you off the stable channel. I could go on … Nextcloud desktop has severe quality control issues.


An number of github issues is even a worse metric than CVEs, many people just post wishlist issues there.


An enterprise setup where people share machines, sure. There are plenty of reasons to be afraid [and mitigations], no need to find them.

Either take control or sell/outsource it, no skin off my teeth. I was replying to someone making the case for 'just trust Google/whoever, lol'

My point is this nears hysterical fearmongering. I'd prefer if you don't go on, but it's more for your benefit.

Stopping before I start my own rant about risk tolerance


Yeah, one CVE is literally "You can use the MacOS variant of LD_PRELOAD on the client to hook libc calls! Oh no!!" This is a bogus CVE; any application can perform arbitrary actions when its system calls are hooked, but it requires such a strong threat model that the adversary realistically gains no ground by doing so.

("A code injection in Nextcloud Desktop Client for macOS allowed to load arbitrary code when starting the client with DYLD_INSERT_LIBRARIES set in the enviroment")


You will note that the PR strengthens that model regardless.


Yeah, it's strange to me that's a CVE. That seems like "working as intended" if I, the owner of the machine, want to load other libraries, why shouldn't it respect that?


Your right of course. No way an individual can compete with an army of specialists.

But for some of us it is a bit of a hobby to run our own infrastructure. And some of it only ever runs on a private network.

I rolled my own docker setup for Nextcloud a few years ago, and couldn't be happier with the outcome. It does require me to log in and update the system and setup from time to time, but that's just a good time to drink a hot bevarage and listen to podcasts in my mind.

For anyone hosting their own instance, Nextcloud offers this scan[0] of your public facing url which might come up with something worth fixing.

[0] https://scan.nextcloud.com/


> I’m confident that the thousands of AWS/GCP/Azure/iCloud security engineers are all doing a more thorough job than you can

I'm not so confident about that:

https://cve.mitre.org/cgi-bin/cvekey.cgi?keyword=azure

It really depends on what you self-host.

https://cve.mitre.org/cgi-bin/cvekey.cgi?keyword=syncthing

"Do everything" solutions go against the principle of minimizing the attack surface.

EDIT: More is not always better in security. With more people doing more things, the statistical odds of miscommunication and misconfiguration increases.


I would dig a bit more into the breakdown of the CVEs:

https://www.cvedetails.com/product/34622/Nextcloud-Nextcloud...

As well as if this reflects a systemic issue with the codebase or if it is just getting much needed attention from security researchers. More CVEs can just mean they're cleaning up after vulns really well. But at the same time, if they have critical vulns over and over again, that might indicate bad coding practices or carelessness.


Nextcloud is well known for it shitty legacy PHP codebase.


Agreed. The breakdown is indeed pretty poor IIRC.

Generally you use these disclosures to make directional decisions about infrastructure. The list of fixed and disclosed CVEs combined with the legacy PHP code base doesn’t really pass the security sniff test. You really wouldn’t know for sure without doing a full code audit.


>A common misconception IMO is that running and owning your own infrastructure is somehow more secure.

If done properly cve-s don't matter that much. You create a headscale install on a pi and the headscale port and your router's ssh (key only) are the only things visible from the outside. Take any other than a home router - aka something with support. And you are done.


> If done properly cve-s don't matter that much.

I think it depends on the CVEs and where they are. If it's a software vuln that requires root or some other complex prerequisites then w/e. But, if we're talking about low level problems in either the OS or network layer (e.g. firewalls, routers) then big clouds are most likely going to have that patched and rolled out more quickly IMO.


Or go with Wireguard. It uses UDP, it has a silent protocol, no one from the outside can see it. (unless they can MITM you)


headscale boils down to syntaxis sugar over wireguard.

To the headscale and tailscale teams members reading this - please don't kill me. You are making awesome things.


> thousands of AWS/GCP/Azure/iCloud security engineers are all doing a more thorough job than you

All these cloud services are just attack surfaces with a huge target on their backs. And the security engineers slip up too [0], in the case of Microsoft it's become more of a meme now. The North Korean hackers basically own them.

[0] https://www.techspot.com/news/102573-microsoft-left-server-c...


Somewhat depends on your threat model. The relative value of an iCloud/aws/gDrive 0day is going to be higher than Nextcloud. If you’re in the category of people concerned about this type of breach, self-hosting a PHP web app and claiming it’s somehow safer wont save you either. For this risky population, neither solution works since attackers are willing to throw expensive exploits at your data in either scenario.

If you aren’t being specifically targeted, then you would care about low hanging fruits discovered by something like automated scanning. Not exposing your service to the internet does solve this assuming you’re confident in the stack which provides this isolation. But managing this stack and performing risk calculus here is actually where the security horse trading happens. I think most people aren’t safer managing this themselves — arguably they’re actually worse off.

I have high standards for the confidentiality of my data. I care about things like lateral movement and the massive attack surface that isolation tech to prevent such movement has. I also won’t design monitoring and alerting, ensure a patch state, or perform code audits on Nextcloud and all the isolation tech required to secure it to a comporable level of security. Because of this, I instead reason around the cost of exploitation. I want it to be higher than what I believe Nextcloud provides and I’d rather require an attacker to use an expensive 0day to extract my data off a cloud provider like Google versus a potentially cheap one against my own infra.


Those cloud specialists aren't configuring your ACLs and firewalls - the most common source of severe security problems in the cloud.


Yes, there is a concept of "shared responsibility" in the cloud. Obviously the provider is going to handle some things and you have to take care of others.


Yes, and those "others" are massively more complex and hard to get right than on non-cloud deployments.


I'm not a security specialist, but it seems to me that while managed services typically have better security and sysadmin resourcing, they also have the downside that their security can fail at a massive scale. If someone defeated the security of, say, GitHub, they could leak all the private repos stored there.

Managed services also have to accept connections from the public Internet, which on-premises solutions do not.


> If someone defeated the security of, say, GitHub, they could leak all the private repos stored there.

In theory you are correct but this is like saying keeping your money under your mattress is safer than a bank.

Yes, in theory someone can still all the money at a bank but the bank is infinitely more qualified/competent to not get robbed than you would be.


banks aren't safe because they're unrobbable, they get robbed all the time. They're safe because they're the ones taking on the risk. Data isn't fungible like cash though.


Running and owning your own infrastructure exposed to the outside world can be more insecure, running your own infrastructure at home in segmented networks with wireguard will solve most problems.


I think the threat model isn't that these popular services are going to be attacked, but that they will engage in the denial of service themselves without legal recourse.


Secure from who I guess.

Like sure, someone particularly interested in your home nextcloud instance could probably find their way in eventually.

But if you are concerned more about dropbox killing your account due to nonpayment, cloud backups getting encrypted and the master key being lost, cloud engineers snooping on your files, cloud platforms targeting ads based on your downloaded files etc etc etc it offers an alternative.


It's written in PHP, the most insecure web language on the planet.


This is a very very old opinion, and not true for years.


What, in your opinion, makes PHP less secure than Python or JavaScript?


Not gp but I'd say because it's littered with footguns like this:

https://dev.to/klnjmm/be-careful-about-the-switch-statement-...



Well that solves the eternal rivalry between ICOM and Yaesu.


Yaesu radio explodes when I throw it at the wall because the UI is so frustrating.


Ok I'll give you that. My 818 is horrid.


I'm impressed that Hezbollah was not using Baofeng.


I think they are better funded than Russia :)


At least they get to keep their two days of WFH.

New policy - five days RTO, two days WFH, per week.


Which LP would this policy most align to? and is there an LP that can enforce the 996?


Frugality. Your cost per hour is reduced.

Learn and be curious. If you’re working on the weekend, you get to do this 24x7.


I chuckle but this may not be far off?


I've had a disturbingly large number of friends cancel on weekend or late evening plans because something at Amazon broke, and they had to drive into a downtown office to fix it.


You mean Amazon can't afford on call support engineers and has to rely on free out of hours labour from its employees? That's astonishing.


I noticed this too and I am afraid they wont ever release this on the web app since they seem to have the best lane and road mapping data.


It does. You can tax loss harvest from companies that you own, so it makes sense to have a bunch of bets that are money pits.


I think the title is missing a word:

“Alphabet _has to_ invest another $5B into Waymo”.

I’d like to remind everyone that it’s still a taxi business with taxi margins (best case).


If you aren't paying drivers, that's a lot of margin, yeah? If you're the only driverless game in town all of your vehicles can potentially earn an the salary of a driver, like 20$ an hour. They'd pay for themselves within a year pretty easily.


There are a ton of operational costs that Uber/Lyft have lobbied to give to cheap contractors with no benefits.


Sure, Uber/Lyft have thus far been successful in claiming a really big share of the margin.

But there's still not too much margin when the driver needs some comp.

Waymo, on the flip side, has a lot of capital equipment added to cars to depreciate, but it's gotten way cheaper and stands to get much cheaper still.


The cars may also last way longer when they’re driving around at 30mph and follow all the traffic laws to the letter. Humans are very hard on cars.


Can it? You still need a ton of very expensive infra, engineering, ops to manage all of this.

Waymo has clearly been playing the “but it’ll be cheaper” game for a long, long time now. It makes sense if you squint really hard and fudge some numbers about unit economics.

If the expected outcome is for it to be eventually cheaper, then why when I open the app does it costs almost twice as much as an Uber or Lyft? You’d think they would want to at least convince investors and train customers that the savings are real and they’re going to prove it by passing it on.


Waymo obviously has nearly no economies of scale yet. We are just entering that part of the regime and it is going to be 2-3 years more before we are spreading those costs across a large fleet.

For me, costs have been comparable to Lyft.


It's common (at least nowadays, in my experience) in English to omit the verb or part of it in headlines.


It’s not a grammar fix, it’s a sentiment fix.


They could have let it die. They didn't have to.


Geez, shelving the best self-driving system, one they've put millions of dollars of R&D into, that would be a god damn shame.


Sunk cost fallacy


I think the companies cited in this article might be weird to compare.

Apple is very RTO heavy because they’re an old school hardware company. Hardware work is easy to demand in office work because: (1) apple secrecy and prevention of leaks and (2) access to lab equipment. #2 likely holds true for spaceX as well.

Adding Microsoft to the mix is weird as nobody I know there actually RTOs.

I think people need to actually specifically measure which roles (senior? engineering?) in tech we are discussing RTO about here. I agree that for most software engineering it backfired. But if you’re an apple hardware engineer, there aren’t many places in town that’ll pay you as much so you’ll accept whatever horrible RTO hand you’re dealt. Companies apply these rules to everyone which is very, very stupid IMO.

I think the most interesting part about this being on the inside is the rationale behind RTO. It’s always the same citing culture, collaboration, or other fuzzy things. It is never quantitative. Are you telling me that the people making these decisions are doing so without data? I think that’s unlikely, it’s just that the data isn’t in their favor and execs are smart enough than to let remote versus not remote become yet another bargaining chip for an employee, let alone senior ones.

TLDR, I think senior vs not senior in tech is likely too much of a generalization. But the people with the actual data aren’t speaking up probably because discussing the results don’t benefit them.


Looks like their NPU (aka ANE) takes up about 1/3 of the die area of the GPU.

Would be interesting to see how much they’re _actually_ utilizing the NPU versus their GPU for AI workloads.


CHERI vs MTE is a bit of a nuanced topic. At least one part of the limiting factor for MTE is that you get a finite number of tag “color codes” which opens the opportunity for some form of probabilistic attacks. Of course this helps with defense in depth as it’s yet another layer of security, but it isn’t as strong of a prevention as a CHERI capability for example.

This page explains it pretty well: https://msrc.microsoft.com/blog/2022/01/an_armful_of_cheris/


This survey by the same author also provides a good comparison https://saaramar.github.io/memory_safety_blogpost_2022/


I eventually need to publish an article about how to run an HSM backed root CA on the cheap with m of n auth.

Using nitrokey and some glue scripts you can get the cost below $500. If anyone is interested, let me know.


I've just started down that route. I've got the nitro key hsm2 in the mail, have heard the advice on using two levels (first root in the Key, and intermediary on the Device for easier revoking). I mainly want to issue client certificates so that I can expose internal sites on the public Internet via proxy without having to require a VPN for all of my users, though I'm also interested in certificate based SSH


Yes, please! I would be interested. Currently i'm fiddling around with vault as an ICA, so this sounds like a good next step


+1 sounds like an interesting read


Please do


an article like that would be great!


Guidelines | FAQ | Lists | API | Security | Legal | Apply to YC | Contact

Search: