Abhishek Tiwari

Abhishek Tiwari, Sascha Groß, Christian Hammer

Android apps cooperate through message passing via intents. However, when apps do not have identical sets of privileges inter-app communication (IAC) can accidentally or maliciously be misused, e.g., to leak sensitive information contrary to users expectations. Recent research considered static program analysis to detect dangerous data leaks due to inter-component communication (ICC) or IAC, but suffers from shortcomings with respect to precision, soundness, and scalability. To solve these issues we propose a novel approach for static ICC/IAC analysis. We perform a fixed-point iteration of ICC/IAC summary information to precisely resolve intent communication with more than two apps involved. We integrate these results with information flows generated by a baseline (i.e. not considering intents) information flow analysis, and resolve if sensitive data is flowing (transitively) through components/apps in order to be ultimately leaked. Our main contribution is the first fully automatic sound and precise ICC/IAC information flow analysis that is scalable for realistic apps due to modularity, avoiding combinatorial explosion: Our approach determines communicating apps using short summaries rather than inlining intent calls, which often requires simultaneously analyzing all tuples of apps. We evaluated our tool IIFA in terms of scalability, precision, and recall. Using benchmarks we establish that precision and recall of our algorithm are considerably better than prominent state-of-the-art analyses for IAC. But foremost, applied to the 90 most popular applications from the Google Playstore, IIFA demonstrated its scalability to a large corpus of real-world apps. IIFA reports 62 problematic ICC-/IAC-related information flows via two or more apps/components.

Bojian Zheng, Abhishek Tiwari, Nandita Vijaykumar et al.

The Long-Short-Term-Memory Recurrent Neural Networks (LSTM RNNs) are a popular class of machine learning models for analyzing sequential data. Their training on modern GPUs, however, is limited by the GPU memory capacity. Our profiling results of the LSTM RNN-based Neural Machine Translation (NMT) model reveal that feature maps of the attention and RNN layers form the memory bottleneck and runtime is unevenly distributed across different layers when training on GPUs. Based on these two observations, we propose to recompute the feature maps rather than stashing them persistently in the GPU memory. While the idea of feature map recomputation has been considered before, existing solutions fail to deliver satisfactory footprint reduction, as they do not address two key challenges. For each feature map recomputation to be effective and efficient, its effect on (1) the total memory footprint, and (2) the total execution time has to be carefully estimated. To this end, we propose *Echo*, a new compiler-based optimization scheme that addresses the first challenge with a practical mechanism that estimates the memory benefits of recomputation over the entire computation graph, and the second challenge by non-conservatively estimating the recomputation overhead leveraging layer specifics. *Echo* reduces the GPU memory footprint automatically and transparently without any changes required to the training source code, and is effective for models beyond LSTM RNNs. We evaluate *Echo* on numerous state-of-the-art machine learning workloads on real systems with modern GPUs and observe footprint reduction ratios of 1.89X on average and 3.13X maximum. Such reduction can be converted into faster training with a larger batch size, savings in GPU energy consumption (e.g., training with one GPU as fast as with four), and/or an increase in the maximum number of layers under the same GPU memory budget.