We present CRYPTFLOW, a first of its kind system that converts TensorFlow inference code into Secure Multi-party Computation (MPC) protocols at the push of a button. To do this, we build three components. Our first component, Athos, is an end-to-end compiler from TensorFlow to a variety of semihonest MPC protocols. The second component, Porthos, is an improved semi-honest 3-party protocol that provides significant speedups for TensorFlow like applications. Finally, to provide malicious secure MPC protocols, our third component, Aramis, is a novel technique that uses hardware with integrity guarantees to convert any semi-honest MPC protocol into an MPC protocol that provides malicious security. The malicious security of the protocols output by Aramis relies on integrity of the hardware and semi-honest security of MPC. Moreover, our system matches the inference accuracy of plaintext TensorFlow.We experimentally demonstrate the power of our system by showing the secure inference of real-world neural networks such as RESNET50 and DENSENET121 over the ImageNet dataset with running times of about 30 seconds for semi-honest security and under two minutes for malicious security. Prior work in the area of secure inference has been limited to semi-honest security of small networks over tiny datasets such as MNIST or CIFAR.

No abstract

New technologies and the ubiquitous use of smartphones have opened the possibilities for more convenient, affordable, fast, and safe options in urban transportation. This has led to the emergence of mobility-on-demand (MoD) systems, such as Uber and Lyft, which aim to provide fast and reliable mobility that is catered to individualistic needs. At the same time, automated vehicle (AV) technology has advanced at an impressive pace. Corporations, such as Google and Tesla (1), have been in a race to develop a fully automated vehicle. The combination of these two promising technologies, known as automated mobility-on-demand (AMoD), has recently attracted interest among both researchers and industry (for example, Uber (2) has started testing AV programs in several states in the US).The term AMoD (3) designates a service similar to MoD or taxi, with the difference that vehicle operations are driverless. AMoD combines the benefits of MoD and AVs in several aspects. First, operational cost is drastically reduced, given the complete removal of driver labor costs and superior energy efficiency of AVs. Furthermore, negative externalities, such as emissions, travel time uncertainty, and accidents, may also reduce, as already observed for MoD (4) and for AVs (5). The latter also observes that AMoD will increase road network utilization, making it possible to transport more passengers with less congestion, with respect to privately owned cars. Fagnant and Kockelman (6) found that AV benefits would amount to between $2,690 and $3,900 annually per vehicle, incorporating decreases in insurance, parking costs, and traffic congestion.It is clear that AMoD is a disruptive technology that will deeply impact the transportation system. Most of the literature (7-11) has focused on the efficiency of AMoD and AVs, in terms of road movement and fleet management. However, only a handful of recent studies have shown the importance 758630T RRXXX10.

No abstract

No abstract

Mobility on demand (MoD) systems have recently emerged as a promising paradigm for sustainable personal urban mobility in cities. In the context of multi-agent simulation technology, the state-of-the-art lacks a platform that captures the dynamics between decentralized driver decision-making and the centralized coordinated decision-making. This work aims to fill this gap by introducing a comprehensive framework that models various facets of MoD, namely heterogeneous MoD driver decision-making and coordinated fleet management within SimMobility, an agent- and activity-based demand model integrated with a dynamic multi-modal network assignment model. To facilitate such a study, we propose an event-based modeling framework. Behavioral models were estimated to characterize the decision-making of drivers using a GPS dataset from a major MoD fleet operator in Singapore. The proposed framework was designed to accommodate behaviors of multiple on-demand services such as traditional MoD, Lyft-like services, and automated MoD (AMoD) services which interact with traffic simulators and a multi-modal transportation network. We demonstrate the benefits of the proposed framework through a large-scale case study in Singapore comparing the fully decentralized traditional MoD with the future AMoD services in a realistic simulation setting. We found that AMoD results in a more efficient service even with increased demand. Parking strategies and fleet sizes will also have an effect on user satisfaction and network performance.

No abstract

Evaluating the impacts of shared automated mobility on-demand services: an activity-based accessibility approach Cite this article as: Bat-hen Nahmias-Biran, Jimi B. Oke, Nishant Kumar, Carlos Lima Azevedo and Moshe Ben-Akiva, Evaluating the impacts of shared automated mobility on-demand services: an activity-based accessibility approach, Transportation