Toolchain

Abstract: In this paper we investigate how a VANET-based traffic information system can overcome the two key problems of strictly limited bandwidth and minimal initial deployment. First, we present a domain specific aggregation scheme in order to minimize the required overall bandwidth. Then we propose a genetic algorithm which is able to identify good positions for static roadside units in order to cope with the highly partitioned nature of a VANET in an early deployment stage. A tailored toolchain allows to optimize the placement with respect to an application-centric objective function, based on travel time savings. By means of simulation we assess the performance of the resulting traffic information system and the optimization strategy.

Abstract: Major cloud operators offer machine learning (ML) as a service, enabling customers who have the data but not ML expertise or infrastructure to train predictive models on this data. Existing ML-as-a-service platforms require users to reveal all training data to the service operator. We design, implement, and evaluate Chiron, a system for privacy-preserving machine learning as a service. First, Chiron conceals the training data from the service operator. Second, in keeping with how many existing ML-as-a-service platforms work, Chiron reveals neither the training algorithm nor the model structure to the user, providing only black-box access to the trained model. Chiron is implemented using SGX enclaves, but SGX alone does not achieve the dual goals of data privacy and model confidentiality. Chiron runs the standard ML training toolchain (including the popular Theano framework and C compiler) in an enclave, but the untrusted model-creation code from the service operator is further confined in a Ryoan sandbox to prevent it from leaking the training data outside the enclave. To support distributed training, Chiron executes multiple concurrent enclaves that exchange model parameters via a parameter server. We evaluate Chiron on popular deep learning models, focusing on benchmark image classification tasks such as CIFAR and ImageNet, and show that its training performance and accuracy of the resulting models are practical for common uses of ML-as-a-service.

Abstract: We investigate a novel approach for image restoration by reinforcement learning. Unlike existing studies that mostly train a single large network for a specialized task, we prepare a toolbox consisting of small-scale convolutional networks of different complexities and specialized in different tasks. Our method, RL-Restore, then learns a policy to select appropriate tools from the toolbox to progressively restore the quality of a corrupted image. We formulate a stepwise reward function proportional to how well the image is restored at each step to learn the action policy. We also devise a joint learning scheme to train the agent and tools for better performance in handling uncertainty. In comparison to conventional human-designed networks, RL-Restore is capable of restoring images corrupted with complex and unknown distortions in a more parameter-efficient manner using the dynamically formed toolchain1.

Abstract: Separation Logic is the twenty-first-century variant of Hoare Logic that permits verification of pointer-manipulating programs. This book covers practical and theoretical aspects of Separation Logic at a level accessible to beginning graduate students interested in software verification. On the practical side it offers an introduction to verification in Hoare and Separation logics, simple case studies for toy languages, and the Verifiable C program logic for the C programming language. On the theoretical side it presents separation algebras as models of separation logics; step-indexed models of higher-order logical features for higher-order programs; indirection theory for constructing step-indexed separation algebras; tree-shares as models for shared ownership; and the semantic construction (and soundness proof) of Verifiable C. In addition, the book covers several aspects of the CompCert verified C compiler, and its connection to foundationally verified software analysis tools. All constructions and proofs are made rigorous and accessible in the Coq developments of the open-source Verified Software Toolchain.