With the emergence of data science, protecting the privacy of individuals’ sensitive information is crucial. In particular, data collection and publication under differential privacy have gained considerable attention in a wide range of fields, such as bioinformatics and machine learning. In this context, the benefit of the concept of smooth sensitivity, which can add tailored noise to respective datasets, has been recognized substantially. However, the existing concept focuses only on absolute values regarding the added noise and does not consider its direction. By adjusting the noise scales in the positive and negative directions for each output element, more tailored and realistic perturbations can reduce overall noise. Therefore, this study proposes a novel concept, direction-oriented smooth sensitivity (DOSS), where the amount of noise is never larger than when using the original smooth sensitivity. Further, there is a lack of theoretical analysis and discussion of the probability distributions used for noise generation; therefore, a concise general form is provided for the first time. We then propose a new \(\epsilon \) -differentially private algorithm using DOSS and our general form, along with efficient computation methods. To demonstrate the effectiveness of the proposed algorithm, we applied it to genomic statistical analysis, which plays a crucial role in the development of personalized medicine. In fact, using DOSS achieved a noise reduction of \(10 \%\) or more. Overall, this study can be an important step toward constructing mechanisms that can add optimal noise even more consistent with reality. The omitted proofs, the Python implementation of our experiments, and supplemental results are available at https://github.com/ay0408/DOSS .