MIT report validates the impact of deep learning on cybersecurity

MIT report validates the impact of deep learning on cybersecurity

There are a lot of buzzwords in the world of cybersecurity marketing. When an emerging concept reaches a certain viral tipping point, it suddenly seems like all the vendors are using the same buzzword, which makes everything more confusing. Artificial intelligence and machine learning are ubiquitous in cybersecurity marketing and often confused with each other and with deep learning. A recent report from MIT clarifies the distinction between the three and highlights the value of deep learning for more effective cybersecurity.

The MIT Technology Review Insights report, “Deep Learning Delivers Proactive Cyber ​​Defense,” is sponsored by Deep Instinct, a cybersecurity vendor that developed the first and only purpose-built deep learning cybersecurity framework. effect. The company, which announced a leadership change this week with Lane Bess, former CEO of Palo Alto Networks and COO of Zscaler, taking over as CEO and Guy Caspi, co-founder and former CEO of Deep Instinct succeeding Bess as Chairman of the Board and transitioning to the role of Chief Product Officer – is on a mission to demonstrate that prevention is better than detection and response, and that deep learning is the differentiator that makes it possible.

Karen Crowley, Director of Solutions Marketing for Deep Instinct, said, “This article from MIT is important to the industry because it explains the key differences between machine learning and deep learning. There is a perception that any AI [artificial intelligence] is equal, and organizations need to understand the differences in the results they can achieve. Deep learning provides a revolutionary methodology to prevent attacks before detection and response. »

Artificial Intelligence vs. Machine Learning vs. Deep Learning

The MIT report explains: “The terms ‘AI’, ‘machine learning’ and ‘deep learning’ are often confused. The technologies are distinct but related. AI is a broad set that encompasses a number of technologies, including machine learning and deep learning. Machine learning is a subset of AI, and deep learning is a subset of machine learning.

In other words, it all falls under the term “artificial intelligence” and tries to simulate human intelligence or problem solving in some way. Machine learning takes this a step further with a model that can learn and improve based on additional data. Deep learning takes machine learning to another level, adding a layered neural network capable of working with exponentially larger volumes of structured and unstructured data to process and learn at a much higher scale.

Prevention and proactive cybersecurity

It is important to understand the differences and not just assume that all AIs are created equally, because when it comes to cybersecurity, deep learning is able to provide benefits that the other two cannot match.

Much of the difference comes down to the data and how the different models are trained. Machine learning typically trains on about 2% of the data, focusing on things like headers and metadata. In contrast, deep learning absorbs 100% of the raw data.

The deep learning model ingests both what good data looks like and what bad data looks like, and it does so on an exponentially larger scale. Millions and millions of samples are passed to the neural network, allowing the model to have better context and greater accuracy to be able to predict behavior and proactively recognize threats with very few false positives.

Deep learning has proven particularly effective in the fight against ransomware. Once the ransomware payload is executed and a victim’s data is encrypted, it’s essentially too late. Sensing and reacting to that moment will do you no good. You need to be able to prevent ransomware encryption in the first place. Deep learning enables the model to understand the DNA of an attack and accurately predict suspicious and malicious behavior. He doesn’t need to have seen that specific attack before, and he doesn’t need to have a full understanding or signature of how the attack works or expect the attack to follow a prescribed scenario. The ability to predict and prevent ransomware attacks before they happen is crucial.

“Deep learning is essential for cybersecurity to stay ahead of attacks such as ransomware,” said Mirel Sehic, global director of cybersecurity at Honeywell. “We need to beat attackers at their own game. Deep learning provides this opportunity by understanding the DNA of files and immediately determining if there is malicious intent before it can land and infiltrate. an environment.

Prove the point

Deep Instinct understands that there is a lot of confusion and misinformation to contend with, both for deep learning versus other AI models, and for the concept of prevention versus the prevailing mantra of detection and response.

This MIT report is one example of Deep Instinct working to demonstrate the value of deep learning and educate the market, but it’s not the first. Deep Instinct also recently engaged with Unit 221B to conduct an extensive, independent test to assess their threat prevention capabilities.

Deep Instinct passed this assessment with flying colors and transformed Unit 221B CEO Lance James from skeptic to believer. The Unit 221B team threw everything they had at Deep Instinct, including custom ransomware using proprietary techniques, and Deep Instinct shut them all down.

Take a look at the MIT report and Unit 221B assessment results and decide for yourself. Perhaps deep learning can break the supposed breach mentality and help organizations actively prevent cyberattacks rather than just trying to detect and respond to them faster.

#MIT #report #validates #impact #deep #learning #cybersecurity

Leave a Comment

Your email address will not be published.