Hacker News Board Hacker News, browsed like a classic message board
Live data from Hacker News

I ditched OpenClaw and built a more secure AI agent (Blink and Mac Mini)

coder.com

Post reply on HN
Print view View original

51–60 of 63 posts

Re: I ditched OpenClaw and built a more secure AI agent (Blink and Mac Mini)

#51

Post by sathish316 »

There are several security flaws in OpenClaw:

1. Prompt injection - this is unsolvable until LLMs can differentiate command and text

2. The bot can leak secrets. The less secrets, API keys, passwords you provide the more useless it is

3. The VM on which it runs can get compromised resulting in leaking private conversations or confidential data like keys. This can be fixed with private VPNs and a security hardened VM or a MacMini like disconnected device.

I’ve found an interesting solution to problems #2 and #3 using a Secure vault, but none so far for Prompt injection. It follows the principle of least privilege, giving secure key access to only the shell scripts that are executed by a skill, along with granting access to the vault for smaller intervals like 15 mins and revoking the access automatically with TTL or time-scoped vault tokens. More details here - https://x.com/sathish316/status/2019496552419717390?s=46

Re: I ditched OpenClaw and built a more secure AI agent (Blink and Mac Mini)

#52

Post by suhputt »

embedding-shape wrote: post #25
suhputt wrote: post #10

so, ignoring the the fact that you yourself didn't actually write this (based on commit history), and the fact that your claims about better security are dubious at best, the most interesting thing I find about this whole situation is - how did you get this to the hackernews front page so fast? that's the real (not-so) secret sauce here :)

> how did you get this to the hackernews front page so fast? Fast? Posted one hour ago. Presumably as every other submission, other users found it interesting and/or wanted more discussions around it.

There were no comments in this thread besides the author's when this made it to the front page. This was very obviously artificially promoted (probably via bots) to the front page - this has become a very noticeable pattern in recent times.

Re: I ditched OpenClaw and built a more secure AI agent (Blink and Mac Mini)

#53

Post by suhputt »

rob wrote: post #20
suhputt wrote: post #10

so, ignoring the the fact that you yourself didn't actually write this (based on commit history), and the fact that your claims about better security are dubious at best, the most interesting thing I find about this whole situation is - how did you get this to the hackernews front page so fast? that's the real (not-so) secret sauce here :)

Not saying "ericpaulsen" is a bot, but the account fits the trend I've noticed of other bots on HN recently: they're all accounts created years ago with zero activity and then suddenly a bunch of comments. Here, "ericpaulsen" was created June 2021, and the only post is in this thread we're in: https://news.ycombinator.com/threads?id=ericpaulsen --- Others caught that fit this trend: https://news.ycombinator.com/item?…

yup this is a daily occurrence at this point

Re: I ditched OpenClaw and built a more secure AI agent (Blink and Mac Mini)

#54

Post by ericpaulsen »

suhputt wrote: post #10

so, ignoring the the fact that you yourself didn't actually write this (based on commit history), and the fact that your claims about better security are dubious at best, the most interesting thing I find about this whole situation is - how did you get this to the hackernews front page so fast? that's the real (not-so) secret sauce here :)

Not sure what commit history you are referring to. I did write this, and I built architecture described.

Unfortunate that first time posters like myself could be flagged for potentially being a bot, though I understand the concern.

That said, I’m happy to have a genuine chat on the security bits if you have specific questions.

Re: I ditched OpenClaw and built a more secure AI agent (Blink and Mac Mini)

#55

Post by ericpaulsen »

suhputt wrote: post #52
embedding-shape wrote: post #25

Earlier quoted context omitted.

> how did you get this to the hackernews front page so fast? Fast? Posted one hour ago. Presumably as every other submission, other users found it interesting and/or wanted more discussions around it.

There were no comments in this thread besides the author's when this made it to the front page. This was very obviously artificially promoted (probably via bots) to the front page - this has become a very noticeable pattern in recent times.

If you have any evidence of artificial promotion, please share.

Re: I ditched OpenClaw and built a more secure AI agent (Blink and Mac Mini)

#56

Post by suhputt »

ericpaulsen wrote: post #55
suhputt wrote: post #52

Earlier quoted context omitted.

There were no comments in this thread besides the author's when this made it to the front page. This was very obviously artificially promoted (probably via bots) to the front page - this has become a very noticeable pattern in recent times.

If you have any evidence of artificial promotion, please share.

already have :)

Re: I ditched OpenClaw and built a more secure AI agent (Blink and Mac Mini)

#57

Post by suhputt »

ericpaulsen wrote: post #54
suhputt wrote: post #10

so, ignoring the the fact that you yourself didn't actually write this (based on commit history), and the fact that your claims about better security are dubious at best, the most interesting thing I find about this whole situation is - how did you get this to the hackernews front page so fast? that's the real (not-so) secret sauce here :)

Not sure what commit history you are referring to. I did write this, and I built architecture described. Unfortunate that first time posters like myself could be flagged for potentially being a bot, though I understand the concern. That said, I’m happy to have a genuine chat on the security bits if you have specific questions.

> Not sure what commit history you are referring to.

lol

Re: I ditched OpenClaw and built a more secure AI agent (Blink and Mac Mini)

#58

Post by TZubiri »

mentalgear wrote: post #19
TZubiri wrote: post #11

Earlier quoted context omitted.

Have you used OpenClaw? My experience has been that it doesn't take input from the world, unless you explicitly ask it to. But I guess that isn't too crazy, if you ask it to look at a website, maybe the website has a hidden prompt. I guess that's more of a responsibility of the LLM model in the security model. That said, I don't think the main dev is serious about security, I've listened to the whole Lex Friedman int…

LLMs famously can't separate data from commands (what you mean by input) - that's one of their core security issues. Check simonw's lethal trifacta. Agreed on all the other points !

Ok, but system prompts are weighted differently and their context weighting is different.

Additionally, there's non LLM inputs, for one, parameters, but also just good old code:

OpenClaw is designed to programmatically read their "SOUL" often, and not forget it, user messages can also be repeated and researched. Compare that to some website's code, and while it may find a way to persist or infect the SOUL, it would need to be something specialized.

You have to admit that even if technically not different, there's a huge semantic and probabilistic difference between owner compile time input and non-owner runtime inputs.

Re: I ditched OpenClaw and built a more secure AI agent (Blink and Mac Mini)

#59

Post by TZubiri »

chasd00 wrote: post #32
TZubiri wrote: post #8

Earlier quoted context omitted.

Don't most ISP routers block ports unless you port forward them though? I wouldn't say that the vulnerability in that case was in OpenClaw, but with the router, nowadays it's expected that ports are blocked unless explicitly allowed in the router.

All home routers block all ports by default. How would they know which IP and port to forward traffic to if not for manual configuration? Also, "listening on all interfaces" doesn't matter on a home network, multi-homed devices don't make any sense in a home network unless you're purposely experimenting or playing with things like that yourself. Further, you're going to configure your router to port forward to only o…

>All home routers block all ports by default. How would they know which IP and port to forward traffic to if not for manual configuration?

I might have to recommend brushing up on your IP/TCP

This is what routers do by default. IP is very different from IP. IP is what they use to know how to route, too long to fit in a post, but MAC to IPs form a local ARP table. Can be many nodes long. Ports are what identify a process in a machine, IP level routers don't care at all about the port (unless you are natting) in order to route, the port bytes are forwarded as is and only parsed by the destination (and constructed by the source)

Post reply on HN