Algorithm Development for Cryptographic Deep Neural Network

Army communications rely on a secure, encrypted radio network.  Army sensors implement a universal communication standard to scramble and unscramble messages.  This becomes a vulnerability whenever Army equipment is confiscated from the battlefield.  Adversaries can use a Side Channel Attack (SCA), augmented by a Deep Neural Network (DNN), to decrypt cipher keys.  The most dangerous of these attacks is a SCA that uses the unintended RF emissions. With a sensitivity of -170 dBm at room temperature, Nokomis has developed the worlds most sensitive RF detectors.  Nokomis leads the market in leveraging these emissions to uncover details about the electronic.  For example, Nokomis can detect a single change in a ring oscillator, or spot changes in firmware loads, based on rich information contained within the unintended RF emission.  In this Phase I, Nokomis will execute a DNN-augmented, Side Channel Attack, which leverages the unintended RF emissions to decrypt a cipher.  Nokomis will achieve this by rotating the cipher key and training the DNN to correlate different RF emissions to known keys.  In this manner the DNN will be capable of identifying the cipher key when presented with fresh RF data.  Such a system would be capable of cracking a specific device, with a specific cipher.  Nokomis will then study the anatomy of the DNN which evolves from this process.  Specifically, Nokomis will report on the DNN architecture, common features and how to adapt this network to new devices and encryption algorithms.  In the option period, Nokomis will apply this DNN to a radically different device and determine the success of the DNN against these new architectures.  When Phase II funding is approved, Nokomis will develop a DNN software toolkit, which will allow Army personnel to build DNNs specializing in decryption.  An RF-based Side Channel Attack is the greatest threat to Army encrypted communications, and Nokomis is uniquely positioned to execute on these attacks.