<p>LoRaWAN is emerging as a promising Internet of Things (IoT) technology to tackle industrial and urban challenges. However, in the context of security, a specific attack known as the Energy Depletion Attack (EDA) aims to drain the battery of LoRaWAN end devices until they become unavailable, posing a potential threat to the network. EDAs usually derive from other attacks, such as flooding or jamming attacks. There are also silent attacks, in which a vulnerability in the end device can increase its processing or network listening activity, consuming its energy without generating additional traffic. Current solutions still need to address the detection of multiple EDAs, and silent attacks are still an unexplored field of research. We propose a Lightweight Architecture for the Detection of EDAs (LADE), which implements statistical distance metrics in analyzing energy data from sensors to detect anomalies in their consumption. LADE diverges from traditional approaches that analyze network traffic, focusing on analyzing energy consumption data. Furthermore, its method allows for the detection of multiple EDAs, including silent ones. LADE encompasses the proposition of detection and learning algorithms, providing a distributed and autonomous intrusion detection system. The proposed architecture is evaluated in two distinct simulation setups, in which jamming and compromising device attacks are simulated to validate the proposed method. The presented results reveal a high accuracy rate in detecting EDAs, with an F1 score above 0.95 for specific system settings. In addition, the proposal presented significant energy efficiency, consuming between 0.1 and 4.1% of the device’s energy depending on the proposal setup.</p>