In recent years, the study of neuromorphic devices has received extensive attention. It is becoming an important branch of the development of artificial intelligence technology. At the same time, natural biomaterials have several priorities, such as biodegradability, good biocompatibility, and non-toxicity, and have important value in novel portable intelligent systems. The egg shell membrane (ESM) is a fiber scaffold composed of highly crosslinked collagen, glycoprotein and cysteine-rich eggshell membrane proteins. It has porous morphology, thermal stability, mechanical strength, etc. Therefore, these protein-based fiber membranes have several potential applications, including nanocatalysts, microbial fuel cells, and adsorption of toxic dyes. This study adopts ESM as electrolyte, exhibiting extremely high proton conductivity of about 6.4×10 ^–3S/cm and extremely high electric-double-layer (EDL) capacitance of about 2.8 µF/cm ²at room temperature. Thus, it has extremely strong interfacial EDL electrostatic modulation capability. Then, indium tin oxide EDL transistor is fabricated by using a single step masking processing and magnetron sputtering deposition technology. The device exhibits typical n-type output curves and transfer curves at low operating voltage. In addition, transfer curves are scanned twice. It is observed that the curves approach to each other quite well, indicating the good stabilities. Owing to the extremely strong proton gating effects, the device exhibits excellent electrical performances. Specifically, ON/OFF ratio, mobility and sub-threshold swing are estimated to be about 2.5×10 ⁶, about 3.2 cm ²/(V·s), and about 213 mV/dec, respectively. With the unique interfacial EDL modulation activities of ESM, the transistor can mimic some important synaptic plasticity behaviors, such as excitatory postsynaptic current (EPSC) and paired pulse facilitation (PPF). With the increase of pre-synaptic spike amplitude, the EPSC value increases correspondingly. With the increase of pre-synaptic spike frequency, the EPSC grain increases, indicating the potentials in high-pass synaptic filtering. By loading 64 potentiation spikes and 64 depression spikes, multi-level synaptic weight can be updated, demonstrating potentiation activity and depression activity. Again, with the same potentiation spikes and depression spikes, synaptic weight value curves approach to each other quite well, indicating that the present ESM gated oxide neuromorphic transistor has good stability. Then, an artificial neural network is adopted to perform supervised learning with Modified National Institute of Standards and Technology (MNIST) database. For simulation, a two-layer multilayer perceptron (MLP) neural network with 400 input neurons, 100 hidden neurons and 10 output neurons is adopted. The best recognition accuracy is as high as 92.59%. The proposed ESM gated oxide neuromorphic transistors have certain potentials in low-cost biodegradable neuromorphic systems.