The Cowry game is a board game of imperfect information that originated in India. It is a game of chance where the player's pieces move along a specified path based on the roll of special dice, known as Cowry shells. This paper proposes using the Q-learning algorithm, an off-policy reinforcement learning (RL) algorithm, for the Cowry game. We implemented a Q-learning-based Cowry player, which learns to play by engaging in a series of games against opponent players. The results show that our Q-learning-based Cowry player outperforms the opponents with a significant winning margin, assuming the opponents follow either a random strategy or a hind-most strategy.

An adversarial attack is an intentional attempt to fool or affect a machine learning model by giving deliberately generated input data, often known as adversarial examples. An adversarial attack exploits flaws in the model's decision-making process, resulting in inaccurate or unexpected results. Attacks on deep learning models, especially in computer vision, threaten their reliability. This work introduces a novel adversarial attack technique. This attack effectively targets vulnerable areas by producing imperceptible perturbations localised to particular sections of an input image by utilising the gradient sign idea. The approach drastically lowers the accuracy of deep learning algorithms by applying gradient-based perturbations throughout the image using the sliding window technique. Extensive experiments on state-of-the-art deep learning architecture were conducted to assess the feasibility of the suggested attack. We identified the specific weak points in the input photos by varying several parameters to control the degree of disturbances. The findings indicate that the nose and eyes are among those parts of the face most susceptible to hostile attacks.

Abstract This research explores the vulnerability of selective reincarnation, a concept in Multi-Agent Reinforcement Learning (MARL), in response to observation poisoning attacks. Observation poisoning is an adversarial strategy that subtly manipulates an agent’s observation space, potentially leading to a misdirection in its learning process. The primary aim of this paper is to systematically evaluate the robustness of selective reincarnation in MARL systems against the subtle yet potentially debilitating effects of observation poisoning attacks. Through assessing how manipulated observation data influences MARL agents, we seek to highlight potential vulnerabilities and inform the development of more resilient MARL systems. Our experimental testbed was the widely used HalfCheetah environment, utilizing the Independent Deep Deterministic Policy Gradient algorithm within a cooperative MARL setting. We introduced a series of triggers, namely Gaussian noise addition, observation reversal, random shuffling, and scaling, into the teacher dataset of the MARL system provided to the reincarnating agents of HalfCheetah. Here, the “teacher dataset” refers to the stored experiences from previous training sessions used to accelerate the learning of reincarnating agents in MARL. This approach enabled the observation of these triggers’ significant impact on reincarnation decisions. Specifically, the reversal technique showed the most pronounced negative effect for maximum returns, with an average decrease of 38.08% in Kendall’s tau values across all the agent combinations. With random shuffling, Kendall’s tau values decreased by 17.66%. On the other hand, noise addition and scaling aligned with the original ranking by only 21.42% and 32.66%, respectively. The results, quantified by Kendall’s tau metric, indicate the fragility of the selective reincarnation process under adversarial observation poisoning. Our findings also reveal that vulnerability to observation poisoning varies significantly among different agent combinations, with some exhibiting markedly higher susceptibility than others. This investigation elucidates our understanding of selective reincarnation’s robustness against observation poisoning attacks, which is crucial for developing more secure MARL systems and also for making informed decisions about agent reincarnation.