This article studies the interesting dynamics of a small neural network with three neurons under electromagnetic radiation. The electromagnetic radiation strength can change the number of the equilibrium points in the neural network, which leads to the diversification of the attractor’s trajectory. Thus, the novel multi-style attractors like one-to-two-spiral attractors and one-to-four-scroll attractors can be generated from the neural network stimulated by electromagnetic radiation. Besides, the plentiful dynamical behaviors are observed in the neural network, such as transitional coexisting attractors, hypogenetic attractors, periodic patterns, firing patterns, transient chaos and intermittent chaos. In terms of hardware implementation, we utilize FPGA to digitally implement the constructed neural network model. The experimental verification results are highly consistent with the numerical simulation results. In the aspects of engineering application, we apply it to pseudo-random number generator and image encryption respectively, test the random performance of different chaotic attractors by NIST test suite, describe the image encryption scheme based on neural network and estimate its security performances. The ultimate outcomes demonstrate that the neural network model with chaotic behavior has superior randomness and wonderful security, which is very suitable for engineering applications based on chaos.

本文研究了在电磁辐射下具有三个神经元的小型神经网络的有趣动力学。电磁辐射强度可以改变神经网络中平衡点的数量，从而导致吸引子轨迹的多样化。因此，可以从电磁辐射刺激的神经网络中产生新颖的多式吸引子，如一到二螺旋吸引子和一到四卷吸引子。此外，在神经网络中观察到了丰富的动力学行为，如过渡共存吸引子、后生吸引子、周期性模式、激发模式、瞬态混沌和间歇性混沌。在硬件实现方面，我们利用FPGA对构建的神经网络模型进行数字化实现。实验验证结果与数值模拟结果高度一致。在工程应用方面，我们将其分别应用于伪随机数生成器和图像加密，通过NIST测试套件测试不同混沌吸引子的随机性能，描述基于神经网络的图像加密方案并评估其安全性能。最终结果表明，具有混沌行为的神经网络模型具有优越的随机性和极好的安全性，非常适合基于混沌的工程应用。通过NIST测试套件测试不同混沌吸引子的随机性能，描述基于神经网络的图像加密方案并估计其安全性能。最终结果表明，具有混沌行为的神经网络模型具有优越的随机性和极好的安全性，非常适合基于混沌的工程应用。通过NIST测试套件测试不同混沌吸引子的随机性能，描述基于神经网络的图像加密方案并估计其安全性能。最终结果表明，具有混沌行为的神经网络模型具有优越的随机性和极好的安全性，非常适合基于混沌的工程应用。