The impact of Machine Learning Security on the resilience of Connected Autonomous Vehicle Architectures

未来在于智能城市，在那里我们的技术正在改善和方便市民的日常生活。智慧城市是一个高度互联的环境，包括移动设备、城市周围的传感器和云中的服务。所有这些组件都为大数据分析贡献了数据，大数据分析使用机器学习为智慧城市的公民带来价值。自动驾驶汽车也受益于机器学习的发展，实现了实时感知道路上的危险、分析传感器数据以防止碰撞和优化路线规划等功能。自动驾驶汽车的未来将是互联和合作的，这样它们就可以从周围丰富的信息系统中受益，帮助制定决策。这将在时间效率、资金和最重要的安全改进方面带来好处。然而，机器学习技术的加入也为对手提供了另一种攻击媒介。对机器学习的攻击不仅会影响使用这些车辆的商业和服务，还将对联网的自动驾驶车辆的乘客和行人安全造成严重后果。因此，我们必须了解对机器学习的攻击如何影响联网自动驾驶汽车网络的弹性。通过将互联的自动车辆网络视为一个复杂的动态和互联的机器学习模型系统，我们可以调查攻击可能的级联影响，以帮助架构师了解这些系统中的薄弱领域。

The future lies in smart cities where we have technology improving and facilitating citizens daily lives. Smart cities are a highly connected environment, which includes mobile devices, sensors around the city and services in the cloud. All of these components contribute data to big data analytics that use machine learning to bring value to the citizens of smart cities. Autonomous vehicles have also benefited from the evolution of machine learning, enabling functions such as the real time perception of hazards on the road, analysis of sensor data to prevent collisions and the optimisation of route planning. The future of autonomous vehicles will be connected and cooperative so that they benefit from the rich information systems around them to assist in decision making. This will provide benefits in time efficiency, money and most importantly improvement in safety. Nevertheless, the inclusion of machine learning technologies also provides another attack vector for adversaries.Attacks on machine learning will have serious consequences for connected autonomous vehicles regarding safety of passengers and pedestrians in addition to impacts on businesses and services utilising these vehicles. We therefore must understand how the resilience of the connected autonomous vehicle networks are impacted by attacks on machine learning. By considering connected autonomous vehicle networks as a complex dynamic and connected system of machine learning models, we can investigate the possible cascading impact of attacks to aid architects in understanding the areas of weakness in these systems.