Side-Channel Attack Analysis on In-Memory Computing Architectures

In-memory computing (IMC) systems have great potential for accelerating data-intensive tasks such as deep neural networks (DNNs). As DNN models are generally highly proprietary, the network architectures become valuable targets attacks. In IMC systems, since whole model is mapped on chip and weight memory read can be restricted, pre-mapped acts a ``black box'' users. However, localized stationary data patterns may subject to other this paper, we propose side-channel attack methodology architectures. We show that it possible extract architectural information from power trace measurements without any prior knowledge of network. first developed simulation framework emulate dynamic traces macros. then performed leakage analysis reverse engineer stored layer type, sequence, output channel/feature size convolution kernel Based extracted information, full potentially reconstructed Finally, discuss countermeasures building offer resistance these extraction attack.