Hacker News Reader

Comment by p1llus | Hacker News Reader

One thing I found that I thought was a fun addition, is using eBPF made this even easier. No need to fight with anyone that is using certificate pinning, mTLS or anything else, you just get the raw plaintext data straight of the wire (right before encryption and right after decryption) and works nicely for most of the agents and IDE's.

That will in practice give you everything from telemetry to prompts, and its funny to see just how much some of them collect/run that is not at all related to your own ask..

A handy alternative when certain applications tend to make it harder to apply a MiTM proxy and you can dump it straight into your own scripts/programs to filter out and store it in whichever format you want for more analysis.

View on HN

Out of curiosity: How?

They don't offload TLS to the kernel, do they? Most apps do it in userspace linked against openssl afaik.

Do you patch that lib? If ebpf "just" operates at network/packet level, I don't see how it can do more than Mitmproxy in regard to avoid DH-PFS/Pinning

It’s done via uprobes which allows ebpf to attach to users processes or libraries, basically like ld_preload but built into the kernel. First link that explains it https://blog.quarkslab.com/defeating-ebpf-uprobe-monitoring.... but there are many more links about it and it is a pretty useful tool for debugging in prod environments. (Also first link: https://blog.px.dev/ebpf-function-tracing/) this also the intended use case. px.dev is btw a cncf project which also helps for ebpf debugging in k8s (comes from new relic)

Exactly this :) You can either attach the uprobes directly to the systems OpenSSL to capture anything using it (curl etc), though most of these LLM clients and similar uses their own static compiled BoringSSL, but its very straight forward to find, it does require some knowledge to determine the right place where you attach the uprobe, but after that it can easily be scripted to automatically find the right offsets (as these are usually stripped binaries)

eBPF started as a network tool, but (according to the creators of it) targeting the network was just a ploy to get a foot in the door so they could start expanding it to other things. They didn’t think the idea would be accepted if they tried to do it all at once. So now it works on many parts of the system, not just the network.

Surprised that works, I thought TLS was done entirely in process space.

I think I found it, My first thought was some sort of builtin ssl library backdoor, but it looks like you do some ld.preload shenanigans to inject a eBPF monitor. I am not sure exactly what the BPF brings to the table here. A convenient interface to intercept the accept() syscall?

It is similar but no real shenanigans but rather a common functionality called uprobes. For most of these LLM clients which uses a static compiled OpenSSL or BoringSSL the functions you attach to is also the same overall, making it easily reusable when you know what to do. There has been plenty of srticles over the time on it, quite a fun thing to play around with. I made it work already with most of them for research purposes.

Reply on news.ycombinator.com