Assessment of autoencoders is frequently constrained to average empirical losses, which lack principled and distributional guarantees. Existing methods incorporating optimal transport are training-centric and fail to provide post-hoc certificates for a trained model. In this paper, we derive a high-probability, data-dependent upper bound on the Wasserstein-1 distance between the input and reconstruction distributions. The bound is computed directly from the empirical mean absolute error on a hold-out set, requiring neither retraining nor the use of an optimal transport solver. We provide an explicit finite-sample deviation radius scaling. Moreover, we introduce an analytic attribution framework based on a joint-Gaussian surrogate, which provides interpretable insight into the structure of the reconstruction error. This framework decomposes the expected reconstruction loss into bias, variance, and alignment components in closed form. This enables an auditable procedure for both certifying and diagnosing autoencoder reconstructions.

In this paper, we present a novel approach of symmetrical pruning for lightweight anomaly detectors based on an autoencoder, leveraging the unique encoder-decoder structure of the autoencoder. We develop an efficient network anomaly detector with reduced computational overhead by computing the reconstruction error between hidden activations of an input and its hidden reconstructions and symmetrically pruning nodes with high error values.

In response to the security threats in wireless networks with concurrent device connections, deploying Intrusion Detection Systems (IDS) at the network edge is a promising strategy. However, this approach must take into account the resource constraints incurred by power-limited edge devices, requiring a lightweight solution to IDS. At the same time, the lightweight IDS solution has to minimize performance degradation as higher detection performance is also a key requirement of IDS. In this paper, we design a lightweight autoencoder with explainability, employing the Shapley value to measure unit importance and link importance. This approach can selectively activate only critical components, thereby reducing the complexity for IDS while effectively lowering its performance degradation. We confirm that the proposed algorithm is robust against the harsh sparsity of the autoencoder. Moreover, the sparsity of the proposed lightweight autoencoder can be easily manageable, such that it can be controlled to satisfy the potential constraints of power-limited edge devices. Therefore, the solution is a suitable algorithm for IDS that can be deployed on edge devices in wireless networks.