Johannes Kinder

Johannes Kinder

@jkinder.bsky.social

Professor @ LMU Munich Security, Program Analysis, Machine Learning

214 followers113 following14 posts

🛬 I'm at USENIX Security in Seattle this week, where on Friday at 2pm my former postdoc Tristan Benoit will be presenting our paper "BLens: Contrastive Captioning of Binary Functions using Ensemble Embedding," joint work with Yunru Wang and Moritz Dannehl from my group. Here's the gist:

View full thread

BLens achieves a 42% increase in F1 score on unseen projects, also improving RougeL, Bleu, and VarCLR for function names. Performance metrics tell only part of the story, however, and the paper shows how even seemingly mispredicted functions can provide valuable information to a reverse engineer.

🛬 I'm at USENIX Security in Seattle this week, where on Friday at 2pm my former postdoc Tristan Benoit will be presenting our paper "BLens: Contrastive Captioning of Binary Functions using Ensemble Embedding," joint work with Yunru Wang and Moritz Dannehl from my group. Here's the gist:

View full thread

Our intuition is that naming binary functions is similar to image captioning, with function names and code as two modalities for the same concept. Inspired by the multimodal image-text model GIT, we embed functions into patches using three pre-trained function embeddings (DEXTER, CLAP, PalmTree).

Our model architecture has two main components, COMBO for pretraining and LORD for finetuning and inference. COMBO uses a contrastive captioning loss (inspired by CoCa), whereas LORD uses an MLM objective to optimize the generation of function names while reducing false positives.

🛬 I'm at USENIX Security in Seattle this week, where on Friday at 2pm my former postdoc Tristan Benoit will be presenting our paper "BLens: Contrastive Captioning of Binary Functions using Ensemble Embedding," joint work with Yunru Wang and Moritz Dannehl from my group. Here's the gist:

Binary reversing is hard. One approach to helping is to generate plausible names for binary functions using machine learning models. But state-of-the-art models based on the idea of translating assembly to English struggle with generalizing across projects and may generate misleading names.

Johannes Kinder reposted

DAVE: Open the podbay doors, ChatGPT. CHATGPT: Certainly, Dave, the podbay doors are now open. DAVE: The podbay doors didn't open. CHATGPT: My apologies, Dave, you're right. I thought the podbay doors were open, but they weren't. Now they are. DAVE: I'm still looking at a set of closed podbay doors.

Interesting case where it seems like precise floating point support in SMT solvers would help with generating a working exploit.

Johannes Kinder reposted

They already tried, with MOOCs. We were told we'd need only like 5 universities to produce "the best" content, the rest could shut down and everyone could learn from MOOCs instead. Which showed exactly how much they understood about education, learning, humans, and other things not measured in bits.

Ted McCormick@tedmccormick.bsky.social

in 8-10 years someone from Silicon Valley is going to invent the university, genuinely unaware that it's been done before

Volodymyr Lutchenko of Ukrainian telecom operator KyivStar shares his experience of defending networks in the face of persistent physical and cyber attacks at #mcsc in a chilling call for increased preparedness

Johannes Kinder reposted

If Trump is removing all environmental regulation, how about we undo the "memory safe languages" mandate? I mean, how is mandating "memory safe languages" compatible with freedom of speech anyhow? Enough policing of speech! Raw pointers for raw milk, our youth needs to be hardened by strcpy.

Learned about Futexes while teaching operating systems this year, so it‘s cool to see this work, and of course to see model checking going strong!

afd-icl.bsky.social@afd-icl.bsky.social

Futex is an under-appreciated Linux system call that backs almost everything you do that involves concurrency behind the scenes. HuguesEvrard and I wrote a paper on using model checking to analyse futex-based concurrency primitives. Check it out! doc.ic.ac.uk/~afd/papers/...

Johannes Kinder reposted

I recently saw an amazing Navajo rug at the National Gallery of Art. It looks abstract at first, but it is a detailed representation of the Intel Pentium processor. Called "Replica of a Chip", it was created in 1994 by Marilou Schultz, a Navajo/Diné weaver and math teacher. 1/n

You are up to date.