Anti-nuclei heavier than $\overline{D}$ are unlikely to be formed during cosmic rays (CRs) propagation, as confirmed by the PHOENIX and ALICE collaborations. $\overline{He}$ observations could be related to Dark Matter interactions or to a primordial origin. Dedicated experiments must possess high charge sign discrimination to observe $\overline{He}$ due to the $He$ abundance in CRs. Detector's effects, such as the rigidity resolution and the internal interactions, may lead to misidentifying matter as antimatter, producing a dominant background over rare signal candidates.
In this work, we developed a Monte Carlo simulation to mimic the response of an AMS-02-like detector, identifying several phenomena that misidentify $He$ as $\overline{He}$. The performances of a fully connected neural network, trained over diverse sources of charge sign confusion, are presented.