Elemendar Monthly News Mail 7 – AI For CTI – December / New Year Special!
Tis The Season to look back on the year and forward to the next one. And ? *drumroll* launch our shiny new website! //elemendar.com/
Our new messaging explains why working with us now will begin reducing your risks immediately and deliver a force multiplier to your whole organisation.
So what of 2019?
Growing a company is hard. Not only do you have to alight on a strong idea in the first place, one that is really needed and not just “so what?”, you then have to assemble the right team and build the thing to fill the need. So that’s what we’ve done from 2017 to now.
But then comes an even tougher challenge: how to communicate what we deliver in a way that helps everyone else understand its value to them and turn that into investment, customers and revenues. And in 2019 we achieved all three!
That’s been our journey this year. We will give ourselves a pat on the back and a Christmas drink of steely resolve to do more of the same and make the next year even greater. If you’ve had a similar year or simply tried to achieve something you set out to do, well done!
It’s a circle. Think. Write. Talk. Listen. Think and Write again. Over and over. One step after another and on in to 2020. The 20s are always good!
We Are Hiring!
If you are or know an ML / NLP engineer who might be excited to work with us in this fascinating field then please get in touch. Read our advert here!
https://angel.co/company/elemendar/jobs/499439-ml-nlp-engineer-elemendar-f-t-p-t-contract-up-to-40-80k-or-pro-rata-plus-equity
Elemendar Technical Roundup of 2019
This has been a year of great progress technically. We released 7 versions of our beta product in as many months and are excited to share a preview of what’s coming next year.
May 2019 – V0.0.1 – First fully packaged version using initial NLP models, primarily identifying IOCs, malware and threat actors. Enabled direct integration with customers’ development stacks and first showcased on our demo site.
V0.0.2 – Extracting tables in PDFs for separate processing. This is necessary because tables do not generally contain well written prose, which our models are trained on and require for context.
V0.0.3 – Integrate regular expression extraction for IOCs within tables extracted from PDFs, to cover the large number of IOCs specified in a tabular format in historic reports.
V0.0.4 – Find and discount footers within PDFs and integrate additional models to reduce false positives.
V0.0.5 – Improved checking of STIX2 validation of extracted objects, to ensure that everything provided in our bundles are totally valid in the STIX2 format.
V0.0.6 – 8x faster PDF table processing and 10x faster PDF text extraction
In January we will release V0.0.7, with newly trained models to be integrated and produce even fewer false positives. Following this V0.0.8 will provide more precise information on the type of an identified threat actor, the type of a target and so on. Then 0.0.9 will link to Mitre ATT&CK patterns.
To finish up this festive newsletter I’ve found this Christmas and AI related story from last December which explains just how hard it is to use AI to be funny.
https://www.technologyreview.com/s/612644/we-tried-teaching-an-ai-to-write-christmas-movie-plots-hilarity-ensued-eventually/
Have a fabulous Christmas everyone and we’ll see you in the shiny New Year!
Stay Frosty
-Tristan
