Measuring the popularity and exploring the readiness of Confidential Computing tools.
In order to have a comprehensive data protection and privacy policy, organizations must ensure the confidentiality and integrity of your data in these states: at rest, in use, and in transit. We previously reviewed the ecosystem of tools for protecting data while in use. We identified over thirty companies and solutions that use any number of Confidential Computing technologies for the development of data, analytic, and AI applications.
The purpose of this follow up post is to compare Confidential Computing technologies using an index that measures popularity. As with our previous post on Experiment Tracking and Management tools, we use an index that relies on public data and is modeled after TIOBE’s programming language index. Our index is comprised of the following components:
- Search: We used a subset from TIOBE’s list (Google, Wikipedia, Amazon) and added Reddit, Twitter, and Stack Overflow into the mix.
- Supply (of talent): This component is based on the number of people who have listed a specific Confidential Computing technique as a skill on their LinkedIn profiles.
- Demand (for talent): We examine the number of U.S. job postings from Linkedin and Indeed that mention a specific Confidential Computing technique.
The results are not surprising. Synthetic data is a hot topic that cuts across several large communities including the AI and machine learning crowd, as well those interested in gaming and the Metaverse. Synthetic data is also used in many settings including in software testing and QA, data augmentation, graphic design, and to enhance privacy. As we noted in previous posts, judging by the number of publications, Federated Learning is currently a popular subject among researchers
We also plot the supply and demand sides of the labor market for each technique. These tools are still quite specialized and advanced, so demand and supply remain relatively sparse.
SMPC for Machine Learning and Analytics
It is important to reiterate that the scores above are based on relative popularity rather than actual practical utility. When it comes to tools for machine learning and analytic products and services, we are actually more excited about new tools that rely on secure multiparty computation.
A Secure Multi-Party Computation protocol (SMPC) allows a program to be executed by participants so that the output is revealed only to the desired parties, and that no inputs belonging to other parties will be revealed to participants other than what can be inferred from the outputs.
CipherMode Labs1 recently introduced algorithms and an architecture to process encrypted data that makes SMPC practical and easy to use. CipherCore is an exciting new open source, high performance library that makes SMPC accessible to data teams. It’s aimed at users and teams already familiar with tools like TensorFlow, PyTorch, and JAX. More importantly, CipherCore makes SMPC useful for analytics and machine learning (model inference and model training). Moreover, CipherCore’s architecture includes an intermediate representation layer that makes it easy to switch off of SMPC, should other Confidential Computing tools become more attractive.
Watch this recent excellent free webinar and learn how enterprises are using SMPC and the CipherCore open-source platform to extract insights while keeping their data fully confidential.
The Data Exchange podcast: CipherMode CEO Sadegh Riazi on building tools for privacy-preserving machine learning and analytics based on secure multi-party computation ↓
Related Content:
- Get Ready For Confidential Computing
- A Guide to Data Annotation and Synthetic Data Generation Tools
- Navigate the road to Responsible AI
- Tech companies are gearing up for the Metaverse
- Complete list of previous tech indices
If you enjoyed this post please support our work by encouraging your friends and colleagues to subscribe to our newsletter:
[1] Full Disclosure: Ben Lorica is an investor in Cipher Mode Labs and an advisor to several startups.
[Image: Privacy Forest by Ben Lorica; original photos from Unsplash, via Infogram.]