We're data scientists sharing open research into machine learning, statistical modeling, program analysis, and human-in-the-loop computer network defense.
Very excited to release something we've been working on for a long, long time: SOREL-20M: The @Sophos / @ReversingLabs 20 million sample dataset, including 10M disarmed malware samples, pre-extracted features, and metadata! SAI Blog post here: https://t.co/xyESeAL0rk
1/
Super excited about all the presentation my team @SophosAI has lined up in Vegas this week. If you are able to attend please attend and ask question! We are eager to hear from the community on our work.
https://t.co/Fo78BpoK53
Excited to attend CAMLIS 2021!
https://t.co/SjxGNa3RPh
4 talks from our work:
Risk-mapping the IPv4 space (@tamasV2)
Clustering-supported threat hunting (@awalinsopan)
Detecting exploitation of CPU bugs (@jarvision__)
The SOREL20M benchmark malware dataset (@rharang / E. Rudd)
New and very lucidly written explainer blog post from Sophos AI's Salma Taoufiq describing one of the multiple ML models and rule-based detection layers we use to detect malicious web content at @Sophos. Read it here:
https://t.co/D0OOP5apLA
@daniel_bilar@hillarymsanders Slightly different metrics, but check out work from @rharang and Sophos AI alum @fel_d on the same topic from BlackHat 2018:
https://t.co/Oxk3RQporn
and
https://t.co/C1QJd6xyTN
What is "catastrophic forgetting" in deep learning, why does it matter, and what can you do about it? Let Sophos AI Senior Data Scientist @hillarymsanders break it down for you in our latest blog post!
https://t.co/v5LBzpiWp2
Sometimes a successful scam is just a simple ask away... Here's what to look out for.
Often, in Business Email Compromise (BEC) gift card scams, scammers will pose as someone from the targets’ company with a seemingly simple request.
More from @SophosAI: https://t.co/Q7oty6XJYV
SOREL-20M is the first production-scale #malware research dataset publicly available released with the intent to accelerate research for malware detection via #MachineLearning. Learn more: https://t.co/429osDYzjE w @SophosAI
If you want more, check out:
Our arXiv paper: https://t.co/0L9tVCbhpI
Our @aivillage_dc talk: https://t.co/5eT7G9go4F
Our previous blog post on CATBERT: https://t.co/MitlBxDn2G
It's the holidays, and gift cards can make great presents, but they're also a target for scammers who want a quick payout. Younghoo Lee and @rharang break down how one of these gift card scams works, and how @SophosAI's email model can help catch them! https://t.co/Hrt82rYtsD
👇ML Yara rule generated by @SophosAI's YaraML tool for matching Sunburst / altered Solarwinds PEs; trained using 3 PEs matching FireEye's rule and tested using 3 *other* PEs matching their rule, which it detected successfully ... (https://t.co/u3TVINmJw3) ...
📢 More OSS work from @SophosAI -- this time, YaraML, an open source project for compiling scikit-learn logistic regression and random forest models to Yara for easy model sharing and deployment. A contribution to bridging ML and signatures in the infosec community 👇
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! https://t.co/U82YyNXC3g