Abstract
Embedded systems are integrated in most modern and nano scale systems. They help these systems perform their functions accurately and quickly. Since embedded systems deal with systems with large data, they suffer from multiple types of errors. The asymmetric detector was an early indicator of errors in embedded systems. The asymmetric detector constructed 'self' templates and then matched them with 'non-subjective' samples for a sequence of operations. This construction occurred in the stages of training and testing. The main objective of this paper is to propose an improvement of the functions of the asymmetric detector so it not only matches exact sequences but also determines the sequences with the same characteristics so it can choose the most effective sequence. The proposal is for an algorithm that is merged with the normal incongruity detector. The proposed algorithm calculates the evaluation of the test data and the normal database data for the prediction. It then finds the most similar data between the testing database and the normal database. The results show that the proposed method is more effective than previous efforts in detecting defects in embedded systems by measuring the degree of similarity between the original and the target data.