Deep Instinct

The difference from the classic approach: a network that learns from the file itself

Deep Instinct was founded in 2015 by Guy Caspi, Dr. Eli David and Nadav Maman. Most security products at the time used classic machine learning: a human expert decides in advance which features of a file to measure, and the algorithm learns to classify based on them, a method attackers can learn and get around. Deep Instinct dropped the human definition step and trained a deep neural network directly on raw files, with the aim of blocking a file before it runs rather than detecting it afterwards.

Early milestones: the World Economic Forum, Nvidia and HP

In 2017 the company won the World Economic Forum's "Technology Pioneer" award, and that same year Nvidia joined as an investor. In 2019 it launched a partnership with HP, under which its technology was built into the HP Sure Sense product sold with the company's computers, an unusual commercial endorsement for a young company. According to an official 2020 funding announcement, its earlier investors also included Samsung.

The Series C, February 2020, according to the official announcement

On 12 February 2020, Deep Instinct officially announced a $43 million Series C round, led by Millennium New Horizons with participation from the new investor Unbound, alongside LG Technology Ventures and Nvidia as an existing investor. The round brought its total funding to $100 million. According to the same announcement, in 2019 the company's annual recurring revenue grew by more than 400% and its customer base tripled.

2021 onward: a Series D and a new CEO

In April 2021, the company raised a $100 million Series D led by the investment giant BlackRock. In 2022, according to the official company page, Lane Bess, formerly CEO of Palo Alto Networks and COO of Zscaler, joined as CEO of Deep Instinct, to lead the next generation of its prevention approach.

The product today: DSX and DIANNA

According to the company's website, its current product is built around the Data Security X (DSX) platform, which promises to block more than 99% of unknown attacks in under 20 milliseconds and with less than 0.1% false positives. Alongside it is DIANNA, a generative AI assistant that gives security teams instant analysis of new threats, an extension of the same 2015 prevention approach into the GenAI era.

The debate that was not settled

The debate over whether deep learning is better for cybersecurity has not been settled conclusively: competitors argued that the approach costs more in resources and is less transparent, since it is hard to explain why a deep network blocked a particular file. It is the same transparency question that accompanies deep learning in every field where it makes meaningful decisions.