Marian Gheorghe

In recent years, the application of membrane computing within the field of robotics has been widely investigated, providing a bio-inspired alternative to traditional control paradigms. This paper proposes an approach that combines Enzymatic Numerical P Systems (ENPS) with Non-dominated Sorting Genetic Algorithm (NSGA-II), a well-known multi-objective evolutionary optimisation algorithm, as core components of a real-time obstacle avoidance robot controller. The ENPS is employed as the computational framework due to its inherent parallelism, modular structure and suitability for sensor-driven control, enabling a natural mapping between sensory inputs and motor actions. To enhance real-time adaptability, we design NSGA-II to interact with the robot controller during the simulation, optimising the key parameters of the equations that define the robot behaviour, which are implemented as production functions within the ENPS model instance. This improvement leads not only to passing the testing scenarios used in designing the robot controller, but also to the robot speed and trajectory optimisation. In addition, the paper provides a comparative analysis, employing non-parametric testing to benchmark the proposed NSGA-II-based controller against static-weight alternatives. A comprehensive suite of quantitative performance metrics, including execution time, traveled distance or memory consumption is utilised to validate the superiority of the proposed approach.

Key agreement protocols based on neural synchronization with Tree Parity Machines (TPMs) offer promising security advantages: they do not rely on trapdoor functions, making them resistant to quantum attacks, and they avoid the need for specialized hardware required by quantum-based schemes. Nevertheless, these protocols face a significant vulnerability: the large number of public message exchanges required for synchronization increases the risk that an attacker, acting as a Man-in-the-Middle, can successfully synchronize their own TPMs with those of the legitimate parties and ultimately recover the shared key. Motivated by the need to reduce this risk, we propose a novel probabilistic protocol that enables two parties to securely estimate the size of the shared key during intermediate steps, without revealing any key material. This estimation allows the protocol to terminate as soon as sufficient key material has been established, thereby reducing the number of synchronization rounds and limiting the opportunity for an attacker to synchronize. We integrate our estimation mechanism into a neural key agreement protocol and evaluate its performance and security, demonstrating improved efficiency and enhanced resistance to attacks compared to existing approaches.

Key agreement protocols based on neural synchronization using Tree Parity Machines (TPMs) have two main advantages: their security does not involve trapdoor problems, which may be solved on a quantum computer, and, unlike their quantum counterparts, they do not require special hardware, which is expensive and difficult to manage. In this paper, we present a new protocol for synchronizing two TPMs. Our network construction uses non-binary inputs with positive bounded weights. We analyzed all major known attacks on TPMs and experimentally proved that the percentage of the key recovered by the attacker drops exponentially in the system's parameters. At the same time, the number of rounds of the protocol increases only polynomially.

Although deep learning models have shown promising results in solving problems related to image recognition or natural language processing, they do not match how the biological brain works. Some of the differences include the amount of energy consumed, the way neurons communicate, or the way they learn. To close the gap between artificial neural networks and biological ones, researchers proposed the spiking neural network. Layered Spiking Neural P systems (LSN P systems) are networks of spiking neurons used to solve various classification problems. In this paper, we study the LSN P systems in the context of a federated learning client–server architecture over horizontally partitioned data. We analyze the privacy implications of pre-trained LSN P systems through membership inference attacks. We also perform experiments to assess the performance of an LSN P system trained in the federated learning setup. Our findings suggest that LSN P systems demonstrate higher accuracy and faster convergence compared to federated algorithms based on either perceptron or spiking neural networks.

Computational models used in specifying biological systems represent a complement and become an alternative to more widely used mathematical models. Amongst some of the advantages brought by these computational models, one can mention their executable semantics and mechanistic way of describing biological system phenomena. This short overview report enumerated some of the computational models utilised so far in systems and synthetic biology, the associated analysis and formal verification methods and tools, and a way of facilitating a broader use of this alternative approach.

Spiking Neural P systems are a class of membrane computing models inspired directly by biological neurons. Besides the theoretical progress made in this new computational model, there are also numerous applications of P systems in fields like formal verification, artificial intelligence, or cryptography. Motivated by all the use cases of SN P systems, in this paper, we present a new privacypreserving protocol that enables a client to compute a linear function using an SN P system hosted on a remote server. Our protocol allows the client to use the server to evaluate functions of the form t 1 k + t 2 without revealing t 1 ,t 2 or k and without the server knowing the result. We also present an SN P system to implement any linear function over natural numbers and some security considerations of our protocol in the honest-but-curious security model.