Introduction
Traditional Artificial Neural Networks (ANNs) often rely on simplified models like the Integrate-and-Fire Model (IFM), which typically utilizes a static firing threshold. However, biological neurons exhibit dynamic thresholds that vary based on prior activity. This research introduces a modified pulse-coupled spiking neuron circuit (MPSNC) that incorporates a "memory threshold," more closely mimicking biological behavior and enabling complex chaotic dynamics suitable for secure applications.
The Modified Pulse-Coupled Spiking Neuron Circuit (MPSNC)
The core innovation of this work is the cross-coupled architecture of two single modified spiking neuron circuits (MSNCs). In this design, the firing of one neuron triggers a reset of both the membrane potential and the threshold of the second neuron.
Short-Term Cross Memory
The threshold of a neuron is determined by sampling a reference signal at the precise moment the paired neuron fires. This mechanism creates a short-term "memory" of the system's state, leading to rich, non-linear dynamics. Unlike many theoretical models, this system was verified using a physical prototype built with off-the-shelf components like LM339 comparators and CD4066 switches.
Mathematical Framework and Chaos Generation
To analyze the system's behavior under multiple periodic signals, the researchers extended the definition of the Least Common Multiple (LCM) to rational numbers. This extension allows for the calculation of a common period $p$ for input signals, facilitating the derivation of an iterative map for the "firing phase."
The interaction between these simple neurons with memory thresholds results in period-doubling bifurcations, a hallmark of the transition into chaos. The system demonstrates extreme sensitivity to initial conditions; a variation as small as $10^{-14}$ in a parameter leads to divergent outputs within 20 iterations.
Application in Image Encryption
The chaotic sequences generated by the MPSNC were applied to a high-security image encryption system. Using the standard "Lena.bmp" grayscale image, the researchers tested the algorithm's effectiveness against statistical and brute-force attacks.
Key Results
- Massive Key Space: The system provides a key space of approximately $1.56 \times 10^{43}$, significantly exceeding the $2^{100}$ threshold required for robust protection.
- Statistical Randomness: The iterative firing phases follow a Gaussian distribution, ensuring that the resulting ciphertext is noise-like and resistant to histogram-based statistical analysis.
- Hardware-Software Alignment: Physical circuit results closely matched Multisim simulations, proving the feasibility of hardware-based chaotic encryption.
Conclusion and Future Directions
While the MPSNC successfully generates complex chaos for encryption, the current mathematical derivation is limited to input signals with rational periods. Future research may explore the implications of irrational periods and the implementation of long-term memory mechanisms, such as long-term potentiation (LTP), to further bridge the gap between artificial and biological neural systems. Scaling these circuits into integrated systems remains a key goal for developing dense spiking neural networks.
Summary updated