tamasV retweeted
Want to exchange MLsec models in a readable text format like Yara? Or use them where heavyweight ML frameworks aren't available? Good news! Today Sophos AI is releasing YaraML, a tool for compiling sklearn binary classifiers to Yara for easy deployment! github.com/sophos-ai/yaraml_…
tamasV retweeted
Want to learn more about the base rate estimation method introduced in the Sophos Threat Report? Check out our blog post on the topic!
ai.sophos.com/2020/11/18/how…
tamasV retweeted
If you enjoyed the thread from Sophos Chief Scientist @joshua_saxe on tough questions to ask cybersecurity ML vendors, you might enjoy this look inside how @SophosAI approaches ML projects from @AdarshKyadige! First of a 4-part series.
ai.sophos.com/2020/11/19/ml-…
tamasV retweeted
If you enjoyed our @aivillage_dc presentation on CATBERT, check out our pre-print for all the technical details! Merging the power of Transformers with domain-specific features for state-of-the-art multilingual spam/BEC detection!
arxiv.org/abs/2010.03484
tamasV retweeted
The @ThreatCoalition weekly report is here, with contributions from @SophosAI. Check it out, and be careful out there!
cyberthreatcoalition.org/cov…
tamasV retweeted
This week in the @SophosAI journal club: Ji, Henriques, and Vedaldi on Invariant Information Clustering (arxiv.org/abs/1807.06653). Maximizing mutual information of the classification of pairs of perturbed images turns out to be a powerful un-/self-supervised learning technique!
tamasV retweeted
Ever wanted a crash course in deep learning for security? @Sophos Chief Scientist @joshua_saxe has you covered! Check out his 2018 BlackHat talk, "Deep Neural Networks for Hackers"!
youtube.com/watch?v=fKJ8sTi6…
tamasV retweeted
Social engineering is a really tricky nut to crack with machine learning approaches. Challenge accepted! Check out our latest post on Context-Aware Tiny BERT (CATBERT): using ML+NLP to catch malicious emails!
medium.com/@sophos.ai/detect…


