Deep Instinct is a cybersecurity startup formed by a group of very experienced cybersecurity professionals. They are all PhD’s in computer science (Deep Learning AI) and they recognized today’s cybersecurity challenges and decided to do something different. To find out more we sat down with Guy Caspi, CEO at Deep Instinct:
Q: Guy, how exactly do you utilize AI with your solutions?
A: I would say that any cybersecurity company is using some AI and it is not a big deal, however there are only few that use machine learning successfully and only one company – Deep Instinct’s implemented deep learning framework has been developed from scratch, without a single line of external code or library. It is written in low level C language, and highly optimized for running on both GPU (for training) and CPU (for prediction).
The propriety deep learning framework is considerably faster and more memory efficient than other publicly available deep learning frameworks, and when running on GPUs result is 100x faster in comparison to CPUs, a speedup that is considerably higher than the numbers reported by other frameworks (in most recent benchmark it was 70% faster than Google’s TensorFlow).
Additionally, we have several deep learning related innovations that allow the deep neural network to be condensed such that it can run locally on the devices, whereas other companies that use deep learning typically host it on the cloud due to its large size. The deep learning framework allows for end-to-end training on hundreds of millions of files, from the raw byte values to the final classification (malware/benign), and thus allows us to train models for any file format and any operating systems, in contrast to traditional machine learning which requires years of domain-specific analysis and feature extraction for each file type.
Q: Why is now the time for a technology solution like Deep Instinct?
A: All traditional machine learning cybersecurity solutions are failing to block new zero day attacks.
Currently Deep Instinct is the only cybersecurity company with advanced deep learning capabilities, and due to the extremely high barrier of entry, we expect no other cybersecurity company to be able to use advanced deep learning for the next two to three years. This is due to the difficulty in recruiting deep learning scientists (and not just engineers), difficulty in implementing deep learning frameworks that can train on hundreds of millions of files, and the difficulty in adapting deep learning to cybersecurity which is immensely more complex than applying deep learning to other domains such as computer vision, speech recognition, or text understanding.
For these reasons Deep Instinct was recently recognized by Nvidia as the “most disruptive AI startup”, out of 800 AI and deep learning related startup companies.
Q: You’ve recently announced $32 Million in Series B Round; could you tell us something more?
A: This round will help us to establish our foot print in the market and continue the innovation.