Defining efficient Intrusion Detection Systems (IDSs) have gained a lot of attractions due to the massive increase in different sorts of attacks in Mobile Ad-Hoc Network (MANET). The Mobile Data Analysis (MDA) and Network Security Monitoring put under seruting the traffic patterns on a mobile network to select the most suspicious part. It contains the rules or the patterns which extracted precise network traffic and identified a malicious node on the network. In this paper, we introduced a Stacked auto-encoder approach for enhancing Intrusion Detection Systems; Stacked autoencoder based approach for MANET (Stacked AE-IDS) is a neural network approach in Machine Learning (ML) for reducing correlation. It attempts to model relevant features with high-level and to get a good representation features from Data Correlation by using multiple processing layers. Stacked AE-IDS method tries to reproduce our input where the dimensionality of the input is the same as the of the output with a reducing correlation. Our proposed Deep Learning-based IDS consists of two phases. The output of the auto-encoder is used as the input of the Deep Neural network (DNN) classifier (DNN-IDS). It focuses on Denial of Services (DoS) attack within labeled datasets which are available for intrusion detection and employs the most potential attacks impact routing services in Mobile Network.