As machine learning, artificial intelligence, and algorithm-based decisions proliferate more and more, the small differences between company methods will become secret recipes to be imitated, copied, or stolen. Hiding the operations that make up an algorithm and the the values that result from years of hard work and data collection is the next line of defense for company security.
Using patented (Patent No. 10,289,816) technology, RHYPT applies functional obfuscation to convert multiple types of algorithms into a method that is indistinguishable regardless of the input. The obfuscated method is made up of unique block instruction tensor data (UBITD) pieces that, when chained together end-to-end, perform the same as the original algorithm. After this obfuscation is completed, the underlying algorithm is presented in a format that prevents reverse-engineering.
Once the initial obfuscation has occured, RHYPT uses patented (Patent No. 10,289,816) encryption technology to fully thwart any reverse engineering attempts. As the coefficients of a method are the result of training that may contain output from proprietary or protected data sources, concealing the coefficients not only protects traces of information from being leaked, but also protects the hard work that went into creating the intellectual property. Once encrypted, the UBITD pieces work like a jigsaw puzzle, where they can be put together in any order, but only make sense in a specific one.
A mobile app developer with a succesful image recognition method may enable mobile phone users to access the recognition software, but only with an open data connection. With a RHYPT protected algorithm, the method can be safely downloaded to individual users' devices. By encrypting with a different key each time, attacks attempting to discover and copy the company's work can be prevented.
A small company or university lab, faced with a large volume of incoming data to be processed by a proprietary method, may want to use a cloud computing service. But without extra security, the architecture and coefficients that make up the algorithm would be exposed to any possible attacks on the cloud as well as the hosting service itself. By using RHYPT technology to protect the method, it can be stored on the cloud and used without worrying about any illicit access or theft.