Various economic, social, and cultural factors have contributed to the proliferation of illegal dumps, causing urban image degradation, population health impacts, and soil, air, and water contamination. Scientists developed remote sensing techniques to identify these red spots and thus contribute to their mitigation and control. They recently used these techniques to detect large areas of illegal waste dumping instead of using expensive field monitoring. Artificial intelligence algorithms have been used to process satellite images due to the availability of satellite images and the increase in the processing capacity of computer systems. This work presents the results of a satellite remote-sensing procedure to detect illegal dumps in one hydrographic subbasin in Oaxaca, Mexico, through a supervised land cover classification using a Random Forest classifier. Two hundred and fifty-six control polygons were used to train the classifier. The classification criteria were the twelve bands of the Sentinel 2A satellite images with a spatial resolution of 10x10 meters, the spectral indices NDVI, MNDWI, SAVI, NDBI, BSI, and the surface slope. Google Earth Engine platform was used to process satellite images. There were 288,100 hectares classified in this way: 65.4% classified as vegetation, 31.5% like bare soil, 2.7% was urban soil and the rest was classified as water or garbage. A confusion matrix calculated the accuracy of the model in 0.9517. The model was not able to accurately distinguish between urban soil, bare soil and garbage due to the similarity of their spectral fingerprints. NDVI and SAVI were the most important spectral indices for detecting litter, and those might contribute to building a spectral fingerprint of litter in the future. Poorly classified areas were discarded through photointerpretation work and post-processing. Finally, thirty-two probable illegal dumps were identified, twelve of which were confirmed on the territory.