Finally another Meetup!

This time, Mats Kvarnström and Josefin Scott from Recorded Future will share some of their work and discuss how they use data from their production environment to do further analysis and evaluations. Please see the abstracts below for more detailed descriptions of their talks.

This event will be held online with the possibility for the audience to ask questions. You will receive a link to the live online meetup closer to the event. RSVP to the event at our meetup page: https://www.meetup.com/machine-learning-gbg/events/273799105/

Hope to see you there!

SCHEDULE:
18.00 Welcome
18.05 Model evaluation using production data by Josefin Scott
18.40 Predicting future exploits of cyber vulnerabilities by Mats Kvarnström
19.15 Goodbye

ABSTRACTS:

Model evaluation using production data by Josefin Scott

When developing models we spend a significant amount of time evaluating performance, simulating real distributions and trying to avoid biases. But how do we ensure that we are getting the expected results after release?

While looking at concrete examples, we will discuss evaluation and monitoring of deployed models, what to consider before deployment, and what to do when things don’t go as planned.

----
Predicting future exploits of cyber vulnerabilities by Mats Kvarnström

Much work has been done on detecting cyber attacks and data breaches while they happen, but not enough on preventing them by predicting future threats. I will demonstrate how pattern analysis of textual references and connections between for example tools and malwares enables us to predict what cyber vulnerabilities are most likely to be exploited in the future.

I will also discuss and demonstrate some problems with estimating prediction precision and recall for a process running live, in particular when the time until ground truth is known is random and even censored.