Linkstream – Telegram
Linkstream
173 subscribers
32 photos
3 videos
2 files
898 links
Various links I find interesting. Mostly hardcore tech :) // by @oleksandr_now. See @notatky for the personal stuff
Download Telegram
Federated Learning: Training ML models without having to hoard all data from the users in the cloud first
https://florian.github.io/federated-learning/
It was a long wait, but I'm finally back to writing about my ideas and projects again. Blogging services and social networks come and go, so it's a standalone website this time :)
https://oleksandr.works/2018/08/30/language-engineering/
Oops, your WiFi/Bluetooth chip can inadvertenly broadcast your encryption keys as electromagnetic noise from the CPU gets amplified and transmitted too
http://s3.eurecom.fr/tools/screaming_channels/
Also, hat tip to them for sharing source code and setup details.
Keys for Pektron keyless entry systems (used in Tesla, McLaren and other high-end cars) are copyable within minutes. Tesla at least acknowledges the issue, unlike other vendors 🤦🏼‍♂️ (via @atemerev)
https://www.esat.kuleuven.be/cosic/fast-furious-and-insecure-passive-keyless-entry-and-start-in-modern-supercars/
Hamiltonian Descent Methods: generalization of the momentum

https://arxiv.org/abs/1809.05042
via @karfly
"oops" [TLDR: you can tell libssh "i'm logged in" and you'll be logged in]
https://www.libssh.org/2018/10/16/libssh-0-8-4-and-0-7-6-security-and-bugfix-release/
Orca: Search engine that finds bugs in code!
TLDR: On large projects reduces time to find the offending commit (of hundreds that went into latest and previous releases) 3x on average, 7x median, up to 45x.

https://www.usenix.org/conference/osdi18/presentation/bhagwan
X.org local privilege escalation (most Unix/Linux systems affected; launch X and enjoy superuser rights)

For added fun, combine with recent libssh vuln :)
https://cve.mitre.org/cgi-bin/cvename.cgi?name=CVE-2018-14665
in environments that are only sparsely rewarding, curiosity is a very big deal. also, lazy robots who decide to watch TV instead of exploration :D

https://blog.openai.com/reinforcement-learning-with-prediction-based-rewards/
"Let's remove Quaternions from every 3D Engine", An Interactive Introduction to Rotors from Geometric Algebra
http://marctenbosch.com/quaternions/
beautiful red team detective story // btw, reaching this level of InfoSec requires one hell of an asset inventory
https://threader.app/thread/1063423110513418240
​​when robots get sufficiently smart, they start to avoid work too 🙂
> CycleGAN learns to “hide” information about a source image into the images it generates in a nearly imperceptible, high frequency signal. This trick ensures that the generator can recover the original sample and thus satisfy the cyclic consistency requirement, while the generated image remains realistic
CycleGAN, a Master of Steganography via @vzezin