Adarsh Kyadige

@adarshdk

Data Scientist@Sophos AI. Vegan. Atheist. Hiking, Fitness, Archery and Tennis enthusiast. Recent paper:

Santa Clara, CA
Vrijeme pridruživanja: veljača 2010.

Tweetovi

Blokirali ste korisnika/cu @adarshdk

Jeste li sigurni da želite vidjeti te tweetove? Time nećete deblokirati korisnika/cu @adarshdk

  1. proslijedio/la je Tweet

    Soon you will know what Pickle Rick tastes like.

    Poništi
  2. proslijedio/la je Tweet
    31. sij

    The view from halfway down.

    Poništi
  3. proslijedio/la je Tweet

    before the last part of my show comes out i watned to thank some ppl whove been there for me since the beginning

    Prikaži ovu nit
    Poništi
  4. proslijedio/la je Tweet
    30. sij

    1\ A file seen at "Downloads\svchost.exe" that doesn't *look* like svchost.exe *might* be a problem. Indeed, AI's , and show that a neural net that takes a file's path alongside its contents gets ~30% better detection.

    Prikaži ovu nit
    Poništi
  5. proslijedio/la je Tweet
    20. sij

    Work by , , and myself at , on detecting malicious web content. We scan web text at multiple spatial scales, and then feed the resulting vector into feed forward layers, training the whole model end-to-end.

    Poništi
  6. proslijedio/la je Tweet
    20. sij

    Thank you :). Reiterating that all royalty money (which has been significant, > $20k), goes to the Environmental Defense Fund (), which does science-based advocacy around climate justice issues.

    Poništi
  7. proslijedio/la je Tweet
    18. sij

    I rarely tweet, or share things concerning , but this was "too good" to pass. Reviewer 2's rejection of one of my papers. I am still laughing.

    Prikaži ovu nit
    Poništi
  8. proslijedio/la je Tweet

    r u ready for my last ride

    Poništi
  9. 28. pro 2019.

    When you just finished a rewatch but see this tweet. *Starts again from s1e1*

    Poništi
  10. 26. pro 2019.

    A good time to think about what the big rocks in our lives are!

    Poništi
  11. proslijedio/la je Tweet
    31. srp 2019.

    Hey, want to hear about all the ways I've found to screw up ML/Security projects, so you can avoid my failures? Come listen to "Security data science: Getting the fundamentals right" at Ground Truth track -- 10am Wednesday!

    Poništi
  12. proslijedio/la je Tweet
    31. srp 2019.

    Interested in facial recognition systems? Want to hear a deep dive on how they work, what it takes to implement one, and what the attack surface looks like? Come hear me and talk at ! There will also be a live demo you can play with and try to fool!

    Poništi
  13. proslijedio/la je Tweet
    31. svi 2019.

    No changes to endpoint software/infrastructure are required to deploy this neural network.

    Prikaži ovu nit
    Poništi
  14. proslijedio/la je Tweet
    31. svi 2019.

    Using detections from multiple vendors, from a threat intelligence feed with semantic malware attribute tags and total number of detections, we combine multiple loss functions to minimize during training, resulting in a significantly enhanced neural network.

    Prikaži ovu nit
    Poništi
  15. proslijedio/la je Tweet
    31. svi 2019.

    Sophos’s Data Science team paper accepted to the Usenix Security Symposium 2019: “Auxiliary Loss Optimization for Hypothesis Augmentation” (ALOHA). Deep learning from multiple threat intelligence sources at once leads to better detection!

    Prikaži ovu nit
    Poništi
  16. proslijedio/la je Tweet
    20. svi 2019.

    Our new manuscript (with and ) on combining file path information into a static detection model is out: Learning from Context: Exploiting and Interpreting File Path Information for Better Malware Detection

    Poništi
  17. proslijedio/la je Tweet
    16. svi 2019.

    I am super excited about our new paper "SMART: Semantic Malware Attribute Relevance Tagging" (with , and Tad) . Feedback and comments welcome.

    Poništi
  18. proslijedio/la je Tweet
    15. ožu 2019.

    Amazing work by my collaborators, , using auxiliary loss to boosting malware detection accuracy. Feedback welcome. ALOHA: Auxiliary Loss Optimization for Hypothesis Augmentation

    Poništi

Čini se da učitavanje traje već neko vrijeme.

Twitter je možda preopterećen ili ima kratkotrajnih poteškoća u radu. Pokušajte ponovno ili potražite dodatne informacije u odjeljku Status Twittera.

    Možda bi vam se svidjelo i ovo:

    ·