A Methodology for Root-Causing In-field Attacks on Microfluidic Executions

Recent research on security and trustworthiness of micro-fluidic biochips has exposed several backdoors in their established design flows that can lead to compromises in assay results. This is a serious concern, considering the fact that these biochips are now extensively used for clinical diagnostics in healthcare. In this paper, we propose a novel scheme for root-causing assay manipulation attacks for actuations on digital microfluidic biochips that manifest as errors after execution. In particular, we show how the presence of a functionally correct reaction sequence graph has a significant advantage in the micro-fluidic context for debugging errors resulting out of such attacks. Such a sequence graph is the basis from which the actuation sequence to be implemented on a target Lab-on-chip is synthesized. In this paper, we investigate the possibility of using this sequence graph as a reference model for debugging erroneous reaction executions with respect to the desired output concentration. Our debugging method consists of program slicing with respect to the observable error in the golden implementation. During slicing, we also perform a step-by-step comparison between the slices of the erroneous output with other erroneous and error-free outputs. The reaction steps are then compared to accurately locate the root cause of a given error. In this paper, we consider two different types of assay descriptions, namely (a) unconditional assays, which have a fixed execution path, and (b) conditional assays that alter the execution at runtime depending on the outputs of sensor observations. Experimental results on the Polymerase Chain Reaction (PCR) and Linear Dilution Tree (LDT) and its conditional variant show that our method is able to pinpoint the errors.