Deep neural network learning based tools for embedded systems under side channel attacks
Disclosed is a method for detecting a side-channel attack (SCA) of an electronic target device, which method includes the steps of: a remote sensing EM radiations emanating from the target device; comparing the EM radiations with base line radiations; and identifying anomalies in said EM radiation.
1 . A method for detecting a side-channel attack (SCA) of an electronic target device, comprising:
remote sensing EM radiations emanating from the target device; wherein the EM radiations are selected from the group consisting of visible light, radio wave, microwave and infrared radiation, and wherein the EM radiations are sensed using a bolometer;
collecting EM signals from the bolometer on a graphene network, and feeding the EM signals from the graphene network into a core comprising a GPU configured to run parallel with the target device to train a deep learning network and track and monitor EM performance, wherein the EM signals are collected as inputs to train the deep learning network, wherein the GPU runs parallel with the target device being monitored to inference a target system performance via the deep learning network and emulate a software or hardware architecture of the target device;
comparing said EM radiations with base line radiations; and
identifying differences in the EM radiation as anomalies.
2 . A method for retrieving encrypted data from an electronic target device, comprising
emulating the target device and a similar model device having similar physical characteristics with EM radiation power signals to generate a cycle-by-cycle power consumption trace and side channel traces for the target device and the model device, wherein emulating the target device and a similar model device employs the method of claim 1 , and employing deep neural networks (DNNs) deciphering encryption keys of the target device.
3 . The method of claim 2 , wherein the target device is a cell phone.
4 . The method of claim 1 , wherein the target device comprises firmware.
5 . The method of claim 1 , wherein the target device comprises a drone.