LIVE Q&A: Smoke and Mirrors

First of all, WELCOME and THANK YOU to everyone out there joining us today. 

@TylerM , Security Analyst at Webroot, is with us for the next hour to answer your artificial intelligence and machine learning questions. If you haven’t seen our latest report on the topic, be sure to read our overview here and grab the report here.

With that, we’ll be getting started in a few minutes with some questions that were submitted earlier by a few Webroot customers. 

For our first question (and probably the best place to start), Eric M. wants to know: 

What is the proper definition of AI and ML?

Our next question comes from Martin H.:

To what extent are cyber criminals using AI and ML to develop next generation threats and how are security vendors looking to combat this?

Scifi has the public misconception. We do not have robots with human like brain or equal like intelligence walking around the office advising our executives.  AI is just Machines automatically doing something in place of a human Any kind of automation is considered AI, so we’ve been using AI since the 50’s (technically)  

Machine learning is Artificial Intelligence that is improving its own ability for its designed task

Machine learning right now is for data agregration. It helps find patterns in large data sets. This is really what helps us classify the almost 2 million brand new, never before seen, unique executable files that we encounter each and every day (when we have less than 30 employees in threat research)

Moving right along, Joe R. wants to know about COVID and the cybersecurity industry:  

How has COVID-19 changed the way the industry operates?

From what I understand, there is not many criminals using AI/ML tools right now -probably less than 1%. That could easily change in the future as it gets better and more accessible. The problem is time. These AI and ML models take massive amounts of time to make and train. Criminals are about money and time is money. 

For the security industry, I think many vendors have seen their solutions either automatically adapt and provide greater efficacy - or not - in light of the recent pandemic.  If products and tools are trained to recognize abnormal behaviors and activity, they will recognize the new family of scams directed at susceptible individuals and businesses in recent weeks. If vendors have to make adjustments to the learning or implementation algorithms, it's good that they're doing so, but an indication that their algorithm was probably too "short-sighted.

Teri M. asks:

What is the single most important cyber security tool for any business?

John H. writes in: 

Why are some vendors claiming that their product is NOT definition based, but yet their AI/ML “Brain” is located right on the disk with the install of the product?

Phishing simulation as part of an education suite. Making sure your employees can understand and avoid most phishing tactics while also holding them accountable when they can’t, is going to be a huge bump to your security posture

That is a great question. AI/ML is huge right now in the cyber security field and I can totally understand when people think they are just being thrown buzzwords. 

AI/ML used properly is going to be HUGE not only in data size but also CPU’s crunching the data. I don’t see the logistics or overall effectiveness of a product where the “brain” is on the endpoint machines' hard drive and is fed definitions. AI&ML models are going to be leveraged on massive backend servers so that vendor might deserve another evaluation.

It looks like this will be our final question and it comes from Daniel C.:

Why/how does AI help in the current threat landscape, just justify moving to a product that supports AI?

AI/ML tools can help people do their jobs more effectively now than ever.  Security professionals are always in short supply, and now possibly unavailable or distracted with other pressing concerns.  Businesses are facing unprecedented demands on their networks and people, so any automation is welcome and beneficial. 

For example as malicious payload campaigns change to take advantage of COVID19 threats, our AI&ML models will also adapt and be able to see these change in behaviors. It really helps our researchers catch the low hanging fruit that there is just so much of. 

That concludes our Q&A today.

Thank you to everyone who sent us their questions and stopped by to learn a bit more about artificial intelligence and machine learning. 

Remember, if you haven’t gotten a chance to look over our findings yet, be sure to read our overview here and grab the report here.

Thanks!