Simplifying SAIL Technology
Secure AI Labs has a new, secure data analysis solution to help researchers run models with hospital data normally restricted by privacy laws. No patient data is shared and no algorithms are exposed. By allowing hospitals to utilize their data in a more secure and safe way, researchers get critical data leading towards future cures and solutions. So how does this AI do it all?
Some may be asking how no patient data is shared and, if that’s the case, how researchers can get anything out of it. The data itself is stored at the hospital on a secure enclave similar to what is found in an iPhone. This is a piece of hardware inaccessible by other computers and outside sources, acting like a vault. When a researcher needs to analyze the data, they send their algorithm to the computer which will run the algorithm itself and send the results back to the researcher.
It’s similar to ordering a secret sauce at a sandwich shop, you don’t see it being made and don’t get the ingredients, just the final result. Even if you found a way to peer into one of the cook’s boards, which represents a virtual machine in this case, the fact that the process is federated over multiple machines “cook’s”, means they wouldn’t get the full picture. This same process helps to protect the algorithms, which are some of the most valuable assets of researchers. The second half of this blog will be slightly denser but don’t worry we will reference the secret sauce again.
Third-party trust is the concept that someone can be added to a 2-party relationship to create a trustworthy environment. Similar to having third-party testing, bringing SAIL in as a third non-bias party helps mediate between hospitals and researchers. It’s like SAIL created the recipe for the secret sauce. They can no longer go in and change that recipe, the recipe is marked with a special code and cannot be swapped. In actual application, this is a number that takes up 32 bytes (256-bits). This code is called a hash and lives directly within the CPU of the computer. Since the hash is implemented during production of the hardware, it acts as a key for the usage of the data and algorithms.
Based on this, you can then get a receipt for your purchase at the sandwich shop. Event logs are produced and marked so that both parties have a record of what entered the machine and what left. This is but another assurance given to both parties by SAIL. These logs will make it easy for compliance officers to audit logs and verify them using the hash.
Lastly, some of you may want clarification on what it means to be federated. Federated learning in reference to AI means you are training models using multiple datasets at the same time; this is federated learning. The other use for federated AI is the ability to compute data models over multiple virtual machines, allowing for larger data computation. In addition, the fact that information for one model is spread over multiple machines means there is more security, as hacking into one machine will only provide a small, unusable portion of the information.
This approach to secure data sharing between hospitals and researchers is a game-changer for the healthcare industry. The AI solution provided by SAIL will open new avenues for hospitals and speed up the medical research process. This is not only a great business opportunity but a chance to lead the way for meaningful change in the health-tech industry.
If you have more questions about our technology, please visit https://secureailabs.com/, and follow @secureailabs on Twitter and LinkedIn.
