It's very refreshing to read such clearly written and concise technical reports.
its-summertime 4 hours ago [-]
I always felt bad about the whole site/browser cross integration that chrome and edge and others do. Now I feel really really really bad about it. Can almost see the marketing team's emails about what features to add, and the developers unwantingly writing the code not expecting such things to open so many holes
xnickb 37 minutes ago [-]
I think they very well understand that this code was full of bugs but no one cared. Likely was assigned to some junior who was making an entry into the team/repos as a non-critical task.
And it's hard to judge them: I'd also rather focus on real features.
tveita 2 hours ago [-]
> So even if you’ve never created a User Profile in the browser, just visiting www.microsoft.com is enough for Edge to create one and sign you in automatically.
This is disgusting behaviour, RCE chain or not.
1 hours ago [-]
zhiyan 4 hours ago [-]
human craft is becoming rare these days
Geezus_42 3 hours ago [-]
Has been for a long time.
cubefox 2 hours ago [-]
It's more that LLMs increasingly outperform humans in finding vulnerabilities, so the share of human discoveries goes down.
account42 7 minutes ago [-]
That's downplaying how LLMs discourage humans from thinking themselves by providing an easier "good enough" path.
cubefox 1 minutes ago [-]
That downplaying the massive increase in LLM capabilities relative to the modest decrease in human capabilities.
sharpshadow 3 hours ago [-]
The popup nightmare never stops.
breakingcups 3 hours ago [-]
This is amazing work.
berofeev 3 hours ago [-]
Excellent write up
cyclopeanutopia 4 hours ago [-]
Stopped at
> I won’t go into too much detail in this post, because I think in this era, you can just ask your agents to fill in the details :)
russelg 3 hours ago [-]
If you can understand nuance and context, you would realise the phrase right before that is saying that he's not going to go into details about how navigation itself works in Chrome/Edge, and to watch the presentation from Google itself about those specific details if you are so inclined.
If you then continued on to read the remainder of the article, you would also discover that there was no AI assistance in this research at all.
est 3 hours ago [-]
It's your loss. The whole article is very good written.
ncruces 2 hours ago [-]
And much better for not going into needless detail.
In fact, a problem with LLMs is that they consistently write more than I want to read.
maeln 3 hours ago [-]
The article does actually go pretty in detail about the exploit.
MonkeyClub 3 hours ago [-]
Same; either put in the work to think it through and show us your thought process and conclusions, or just post your prompts and call it a day and everyone gets to move on.
russelg 3 hours ago [-]
If you continued to read the article, you would discover that he indeed goes through his entire thought process in building this chain. Including the fact this was done with no AI assistance at all.
MonkeyClub 2 hours ago [-]
I stand corrected then, I’ll return to it!
tsujamin 3 hours ago [-]
Imagine missing such a great bug-chain because an off-hand AI comment at the start put you off
No memory corruptions, no heap shenanigans, just pure logic bugs, almost made easy to understand for the reader.
As a bonus, they also posted a reverse chain: popping Edge from calc [0].
[0] https://www.youtube.com/watch?v=Bh1zglq4mLE
And it's hard to judge them: I'd also rather focus on real features.
This is disgusting behaviour, RCE chain or not.
> I won’t go into too much detail in this post, because I think in this era, you can just ask your agents to fill in the details :)
If you then continued on to read the remainder of the article, you would also discover that there was no AI assistance in this research at all.
In fact, a problem with LLMs is that they consistently write more than I want to read.