Institution:
We note that many emerging AI-based functionalities are intrinsically stochastic (e.g., camera-based object detection), and hence, their correctness must be judged semantically, with room for variations across correct outcomes (e.g., confidence must be above a given threshold).
Building on this observation, we propose strategies to create DMR and TMR implementations of AI-based functionalities that bring not only fault tolerance against random hardware faults, but also against AI model inaccuracies. Those strategies, which can be realized with software-only means and ported to virtually any computing platform, build on input data modifications affecting the inference computations, but not the expected semantic output (e.g., introducing some limited random noise in the input data). Moreover, we have implemented a tool for image and video processing aimed at facilitating the reproducibility of our evaluation results, and enabling others to use it and conduct further research on input transformations.