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

Skill issue

Jokes aside, I agree on the sense that this should be a solved problem. On the other hand, if the only implementation is a broken one, there's only one way to help yourself, and that is learning how to do it and helping yourself


Being in Kubecon this week and now this project, I'm a bit surprised at how heavily some people lean into Operators. I've always been hesitant to deploy operators into my clusters unless they provide a significant advantage compared to just using argocd just because of how heavily they can hit the k8s API. Not sure if anyone has run into the same issues?


Doesn't it depend on how widely and how many things an operator subscribes to?

I'm pretty invested in vault-secrets-operator and vault-config-operator.

Simply because of the amount of custom scripting I don't have to do.

But I have thought: I wonder if one operator could do the job.

Looking at vault, VSO and VCO codebases, you get a little daunted. There's a lot of code.


This all stinks of Lazarus:

https://en.wikipedia.org/wiki/Lazarus_Group

I've done incident responses for this exact type of attack multiple times. They've gotten much better organized lately and will often contact developers directly (over LinkedIn or WhatsApp) to run this type of attack. (Although, usually pretending to run a test for a job interview -- which is maybe why the author was confused about the code)


Why assume it is Lazarus?

This sort of an attack is comically simple to pull off with a 12b obliterated LLM model and some basic scripts and proxies.

Security has to evolve, or the world will be cooked by script kiddies running email loops.

There's really nothing sophisticated about this these days, and it's only a short matter of time before it becomes commonplace.


Fair challenge, you're right that there's nothing sophisticated about this type of activity, but if you look at Lazarus activity this is their ttp. I mentioned TraderTraitor, go look them up (that sounds terse, it's not meant to be). They stole a couple of hundred million dollars in the past 6 months. They're not particularly sophisticated in terms of ttp, but because they're a nation-state actor they it's an entirely different threat model than script kiddie.

Attribution is hard, but if we're talking about defending, there's little cost to assuming Lazarus-style threat actor.


>Attribution is hard

Very. Frustrating how many "Nation State" attributions boil down entirely because it's some Russian hacker, and they're using some publicly known bit of malware.

No 0days, no magic...


Yeah, the only "sophistication" is the social engineering aspect. Which back in the day a lot of hackers weren't very good at.

But people train up and develop skills.


100%. I can't find it now, but someone last month posted a similar story on HN. The threat actor had stolen someone's GitHub account and altered their otherwise legitimate looking repo. They'll expend a lot of effort in order to masquerade and trick you. TraderTraitor is another good DPRK example.

Anyone reading - if you're ever a victim, worth reporting to your national CERT and your org. The CERT can provide advice, it's useful for their threat intel, and your org can check their systems. You might not be the end target.


Again, you don't need to be a "Nation State" to think this up, and then implement with a few scripts and a bit of LLM.

Some 13yo kid with a VPS could do it


I think this was posted because the of the recent Npm malware fiasco. The malware monkey-patched native JS functions to replace strings that matched crypto addresses in certain the fetch, and XMLHttpRequest functions:

https://www.aikido.dev/blog/npm-debug-and-chalk-packages-com...

Considering there's no way to check whether a function is monkey-patched, this just tells me the JavaScript ecosystem was not designed with malicious actors in mind


It wasn't, but that is not why.


This was discovered last year:

https://www.youtube.com/watch?v=D0st_6sE7Bk


I think the issue with this is that you'll never be able to have the deep integration current FMD implementations have (where you have Bluetooth beacons to track the phone even when it's "off") and also be able to use anything that requires a signed bootloader and OS (such as banking apps)


Not OP and don't work at Jane Street, but I have a friend that does and yes, they do.


I think this article sidesteps what this bill also means: a return to cold-war-era thinking of "passive" advancement of military capabilities for the "just in case" scenario


I wonder if this was cyber sabotage....

Obviously just speculation but considering the escalating tensions between Russia, the EU and the US, I wouldn't be surprised if it were


I work in the security industry and use WIZ and while I do despise all of the buzzword acronyms this industry has come up with, CSPMs have been one of the few tools that have actually made my life significantly easier. Due to the nature of the industry I work in, there is a lot of regulation that we need to comply with, and CSPMs (and wiz in particular) gives us both observability and alerting for all of our resources in our cloud environments, including the configuration of the cloud environments themselves. I don't know how they managed to get a $32B offer so soon after coming out of stealth, but considering the amount of problems it solves for me and my team, I can see why they're doing well financially. We're definitely happy with the pain point the product fixes.

I can now say "I know for a fact we have x number of AWS/GCP/Azure accounts that are either not using our IdP or 2A, here's a list" without having to script across multiple cloud APIs

Similarly, I can say "here's a list of people that accessed x resource in the last y days". It really makes my life easier when I want to access metrics about my company's cloud environments


Is this a difficult problem to solve? There’s only a handful of major cloud players and these questions don’t seem terribly complicated.

Or is it that it lets you answer arbitrary questions of this sort without having to figure out how to get that data?


CSPM is most valuable for large enterprises that have many cloud tenants as they can provide visibility across the entire footprint in one place.

Consider an enterprise that wants to say "list all the cloud storage buckets we own that are not in the US and are publicly readable and have a name containing 'foo'" - and they have several of each of AWS, Azure and GCP organixations because of acquisitions that aren't fully integrated yet.

Wiz answers that in ~5 seconds, with a rich query language and a bunch of prebuilt rules and detections on top of it, including for tracking compliance with various frameworks.


Conceptually, I don't think CSPMs are answering complicated questions, however there's quite a lot of complexity (IMO) in scaling the answers consistently, and keeping up to date with all of the tests that need to be implemented.

If you think about the number of services that AWS/GCP/Azure have, adding good compliance checks across even a portion of those is quite a lot of work :)

A small example from an area I know something about is maintaining the CIS Kubernetes benchmarks (which are used by a lot of CSPM products as a source of rules).

Here you've got the different Kubernetes distributions and then each of the cloud distributions has its own CIS benchmark as the checks are different depending on the cloud in use. Then you have changes over time as different clusters run different versions of Kubernetes, so have different checks. Then you add in that the benchmarks don't release with every new version of Kubernetes, and you can end up with quite a complex matrix of checks.


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

Search: