The global context and the growth in international trade is widely recognized as a boon for the trafficking of illicit materials. In the area of homeland security, the approaches used have failed to provide an up and running technique for non-intrusive on-site inspections. Neutron-induced reactions rely on precise measurement of gamma spectra, which is generally very complex due to high level of background noise. Active photon interrogation methods have also been overlooked but only in the context of actinide detection using photofission reactions (γ,f). This work describes the design of a novel method for the detection of illicit materials based on active photon interrogation associated to photoneutron spectrometry. It is the first attempt to explore the principle of inducing photonuclear reactions on samples to determine the composition of light elements such as nitrogen, oxygen and carbon and therefore to extend the use of active photon interrogation for the detection of conventional explosives, narcotics and chemical weapons. The approach is based on a source of photons produced by electron linear accelerator to induce photonuclear reactions on different materials, on the spectrometry of the photoneutrons created, and on the implementation of a new digital analysis method based on a convolutional neural network for extracting the neutron contribution for each photon energy of the bremsstrahlung spectrum in the total neutron spectra measured. The results of this study suggest that the use of neural networks trained with Monte Carlo simulated spectra can be a digital alternative to the "tagged photon" experimental method. Our study, therefore, provides the foundation of a new way of extracting neutron contributions for discrete photon energy of the continuous bremsstrahlung spectrum.