This week, we announced the launch of Anjuna Enterprise Enclaves—a simple and secure data security solution. While Anjuna Security was founded in 2017, our story actually starts decades earlier.
My co-founder, Yan Michalevsky, and I met when we worked together in Israeli intelligence in the early 2000s. One challenge we often bumped up against was we couldn’t completely protect code or data deployed in insecure or hostile environments. Even when we deployed layers of software protections, there was no way to completely get around the risk of exposure.
Why? Because no matter the protection, critical data such as encryption keys or application code are always eventually exposed in memory and CPU. You simply can’t use data and secure it at the same time. This means that anyone depending on that data remaining private finds themselves in a rather uncomfortable state of data insecurity.
After the army, Yan and I took divergent career paths but both ended up in Silicon Valley.
I went into industry, where at OpenDNS I was again haunted by this issue of inherent data insecurity. There, we held the private keys of customers who used our TLS termination solution. If those keys fell into the wrong hands, not only would our customers be extremely vulnerable to a wide range of attacks, but the OpenDNS reputation would have never recovered. Here again there was little we could do. Even with layers of security, company insiders or bad actors could ultimately retrieve the keys directly from memory. Whoever was able to get access to a computer utilizing those keys could potentially steal them.
It became clear to me this was a significant problem that every organization—either in the private or the public sector—must deal with, and there were no good solutions in the market. I realized the shift to public clouds, would only make the problem larger, because now employees of the public cloud providers would be added to the long list of people with access to enterprise code and data.
Meanwhile, Yan went into academia, studying cybersecurity at Stanford with Prof. Dan Boneh, one of the foremost experts on applied cryptography on the planet. Through the course of his PhD, Yan learned of enclaves and Intel SGX in particular. He collaborated with other researchers on improving the practicality of running applications in hardware-based enclaves .
Yan and I remained in contact, and we often discussed our ongoing frustrations with data security. We realized we were struggling with the same data security problem from different angles, and that the memory exposure issue was at the heart of many enterprise data breaches. For the first time, we had a technology in secure enclaves that would feasibly allow us to do something about this.
For years, we only aspired to solve “data insecurity.” We watched as Apple and others implemented similar technology in cellphones. Finally in 2015, Intel and AMD enhanced their CPUs with a set of secure instructions that encrypt and isolate memory, and allow data to be decrypted only inside the CPU. That was the foundation we needed to solve the data security problem that had been nagging at us for years. Now we could actually do it!
We started Anjuna with one simple vision: All enterprise applications and data should be protected everywhere.
The release of Anjuna Enterprise is an exciting start….and we’re excited to see so many big industry players—from Intel to AMD to Microsoft Azure and others—joining the effort.
Now that you know where we’ve come from, we’d like to invite you to join us moving forward. Over the coming months, we’ll discuss what we know in this blog, and we’ll go into more depth about what this technology means and how it will revolutionize security. We’ll also track new developments in the market to keep you on top of this new wave that may significantly reshape IT. It’s going to be an exciting journey. We hope you’ll join us.
Additional Blog Articles
| 3 MINUTE READ
Protecting Byzantine Fault Tolerance with Trusted Execution
You might be familiar with non-BFT consensus protocols like Paxos and Raft. These protocols can tolerate crash failures in up to 1/2 of the nodes,...