Deep learning based visual sensing has achieved a ractive accuracy but is shown vulnerable to adversarial example a acks. Specically, once the a ackers obtain the deep model, they can construct adversarial examples to mislead the model to yield wrong classication results.Deployable adversarial examples such as small stickers pasted on the road signs and lanes have been shown effective in misleading advanced driver-assistance systems. Many existing countermeasures against adversarial examples build their security on the a ackers' ignorance of the defense mechanisms. us, they fall short of following Kerckho s's principle and can be subverted once the a ackers know the details of the defense. is paper applies the strategy of moving target defense (MTD) to generate multiple new deep models a er system deployment, that will collaboratively detect and thwart adversarial examples. Our MTD design is based on the adversarial examples' minor transferability to models di ering from the one (e.g., the factory-designed model) used for a ack construction. e post-deployment quasi-secret deep models signi cantly increase the bar for the a ackers to construct e ective adversarial examples. We also apply the technique of serial data fusion with early stopping to reduce the inference time by a factor of up to 5 while maintaining the sensing and defense performance. Extensive evaluation based on three datasets including a road sign image database and a GPU-equipped Jetson embedded computing board shows the e ectiveness of our approach.

Damage-induced retraction of axons during traumatic brain injury is believed to play a key role in the disintegration of the neural network and to eventually lead to severe symptoms such as permanent memory loss and emotional disturbances. However, fundamental questions such as how axon retraction progresses and what physical factors govern this process still remain unclear. Here, we report a combined experimental and modeling study to address these questions. Specifically, a sharp atomic force microscope probe was used to transect axons and trigger their retraction in a precisely controlled manner. Interestingly, we showed that the retracting motion of a well-developed axon can be arrested by strong cell-substrate attachment. However, axon retraction was found to be retriggered if a second transection was conducted, albeit with a lower shrinking amplitude. Furthermore, disruption of the actin cytoskeleton or cell-substrate adhesion significantly altered the retracting dynamics of injured axons. Finally, a mathematical model was developed to explain the observed injury response of neural cells in which the retracting motion was assumed to be driven by the pre-tension in the axon and progress against neuron-substrate adhesion as well as the viscous resistance of the cell. Using realistic parameters, model predictions were found to be in good agreement with our observations under a variety of experimental conditions. By revealing the essential physics behind traumatic axon retraction, findings here could provide insights on the development of treatment strategies for axonal injury as well as its possible interplay with other neurodegenerative diseases.

Design of clock synchronization for networked nodes faces a fundamental trade-o between synchronization accuracy and universality for heterogeneous platforms, because a high synchronization accuracy generally requires platform-dependent hardware-level network packet timestamping. is paper presents TouchSync, a new indoor clock synchronization approach for wearables that achieves millisecond accuracy while preserving universality in that it uses standard system calls only, such as reading system clock, sampling sensors, and sending/receiving network messages. e design of TouchSync is driven by a key nding from our extensive measurements that the skin electric potentials (SEPs) induced by powerline radiation are salient, periodic, and synchronous on a same wearer and even across di erent wearers. TouchSync integrates the SEP signal into the universal principle of Network Time Protocol and solves an integer ambiguity problem by fusing the ambiguous results in multiple synchronization rounds to conclude an accurate clock o set between two synchronizing wearables. With our shared code, TouchSync can be readily integrated into any wearable applications. Extensive evaluation based on our Arduino and TinyOS implementations shows that TouchSync's synchronization errors are below 3 and 7 milliseconds on the same wearer and between two wearers 10 kilometers apart, respectively.

This paper presents a robust monocular visual teach-and-repeat (VT&R) navigation system for long-term operation in outdoor environments. The approach leverages deeplearned descriptors to deal with the high illumination variance of the real world. In particular, a tailored self-supervised descriptor, DarkPoint, is proposed for autonomous navigation in outdoor environments. We seamlessly integrate the localisation with control, in which proportional-integral control is used to eliminate the visual error with the pitfall of the unknown depth. Consequently, our approach achieves day-to-night navigation using a single-experience map and is able to repeat complex and fast manoeuvres. To verify our approach, we performed a vast array of navigation experiments in various outdoor environments, where both navigation accuracy and robustness of the proposed system are investigated. The experimental results show that our approach is superior to the baseline method with regards to accuracy and robustness.