Deep learning can be used to classify waveform characteristics (<italic>e.g.</italic>, modulation) with accuracy levels that are hardly attainable with traditional techniques. Recent research has demonstrated that one of the most crucial challenges in wireless deep learning is to counteract the channel action, which may significantly alter the waveform features. The problem is further exacerbated by the fact that deep learning algorithms are hardly re-trainable in real time due to their sheer size. This paper proposes <italic>DeepFIR</italic>, a framework to counteract the channel action in wireless deep learning algorithms <italic>without retraining the underlying deep learning model</italic>. The key intuition is that through the application of a carefully-optimized digital finite input response filter (FIR) at the transmitter’s side, we can apply tiny modifications to the waveform to strengthen its features according to the current channel conditions. We mathematically formulate the <italic>Waveform Optimization Problem</italic> <xref rid="deqnWOP-deqnC1" ref-type="disp-formula">(WOP)</xref> as the problem of finding the optimum FIR to be used on a waveform to improve the classifier’s accuracy. We also propose a data-driven methodology to train the FIRs directly with dataset inputs. We extensively evaluate <italic>DeepFIR</italic> on an experimental testbed of 20 software-defined radios, as well as on two datasets made up by 500 ADS-B devices and by 500 WiFi devices and a 24-class modulation dataset. Experimental results show that our approach (i) increases the accuracy of the radio fingerprinting models by about 35%, 50% and 58%; (ii) decreases an adversary’s accuracy by about 54% when trying to imitate other device’s fingerprints by using their filters; (iii) achieves 27% improvement over the state of the art on a 100-device dataset; (iv) increases by <inline-formula> <tex-math notation="LaTeX">$2\times$ </tex-math></inline-formula> the accuracy of the modulation dataset.