News

Eric Micallef, a senior in the Embedded Systems (EMBS) MSE program, is one of the 2019 recipients of the J.P. Eckert Master’s Fellowship. Congratulations, Eric!

PRECISE members Insup Lee and Oleg Sokolsky, along with Sampath Kannan, Moonzoo Kim, and Mahesh Viswanathan, win the Test-of-Time Award for their 2001 paper titled "Java-MaC: a run-time assurance tool for Java programs" at the 2019 Conference on Runtime Verification.

Luan Nguyen, Gautam Mohan, James Weimer, Oleg Sokolsky, Insup Lee and Rajeev Alur won Best Paper Award on their "Detecting Security Leaks in Hybrid Systems with Information Flow Analysis" at ACM-IEEE International Conference on Formal Methods and Models for System Design (MEMOCODE), 2019.  

Congratulations to Luan and all!

Philly Inquirer article on Uber Elevate quotes our "Fly-by-Logic" research.

Kihong Heo, Mukund Raghothaman (soon to join USC CS), Xujie Si, and Mayur Naik are the recipients of the Distinguished Paper Award at the Conference on Programming Language Design and Implementation (PLDI 2019) for their paper "Continuously Reasoning about Programs using Differential Bayesian Inference".  Their abstract could be found at https://www.youtube.com/watch?v=TebtiCG4HCA

Software developers are struggling to manage the rapidly growing size and complexity of the systems they develop, and are increasingly looking to sophisticated automatic techniques to help them improve the quality of their code. Static analysis is one such popular technique, and promises to automatically find bugs in software before it is even run. Recent years have witnessed the widespread adoption of static analysis tools by industry, from aerospace and safety-critical medical applications to security-conscious cloud service providers.  Unfortunately, the thoroughness of static analyzers is a double-edged sword, and developers are often overwhelmed by the number of warnings produced by tools. In their paper, Dr. Naik, with his postdocs Kihong Heo and Mukund Raghothaman, and his Ph.D.c Xujie Si, developed an AI-based system to prioritize these alarms based on how relevant they are to the developers’ immediate changes. Their system, Drake, will be more effective at identifying vulnerabilities, and promises to eliminate defects as soon as they are introduced into the code-base.  Our PRECISE team plans on deploying Drake on commercial software hosting services such as GitHub, in the coming months.

Two of Rajeev Alur's doctoral students, Konstantinos Mamouras (now at Rice) and Arjun Radhakrishna (now with Microsoft), presented their work at PLDI 2019.  

Links to their papers:

Arun Iyer, Manohar Jonnalagedda, Suresh Parthasarathy, Arjun Radhakrishna, Sriram K. Rajamani: Synthesis and Machine Learning for Heterogeneous Extraction.

Konstantinos Mamouras, Caleb Stanford, Rajeev Alur, Zachary G. Ives, Val Tannen: Data-Trace Types for Distributed Stream Processing Systems.

Linh Thi Xuan Phan has been appointed the co-director of the Data Science (DATS) Master's program. 

Linh is widely known as a dedicated teacher and someone highly engaged with master's students in education and research.  Many DATS students are taking her popular and highly-rated CIS 505, which covers topics related to big data infrastructures and cloud computing. She will add new dimensions to the program, given her expertise in big data infrastructures, cloud computing, and enhance DATS connections to embedded (health) device data analytics.

Congratulations, Linh!

We are pleased to welcome Justin Gottschlich, a respected and well-known leader in the industry, to PRECISE.  He has been appointed Industrial Board Chair and Executive Director of AI R&D.

Meng Xu, Linh Thi Xuan Phan, and Hyon-Young Choi (University of Pennsylvania); Yuhan Lin (Northeastern University); Haoran Li and Chenyang Lu (Washington University in St. Louis); and Insup Lee (University of Pennsylvania) are the recipients of the Best Paper Award at the 25th IEEE Real-Time and Embedded Technology and Applications Symposium (RTAS), part of the Cyber-Physical Systems Week (CPSWeek), that took place in Montreal, Canada on 15-18 April 2019.  Their paper titled “Holistic Resource Allocation for Multicore Real-Time Systems” presents a holistic cache and memory bandwidth resource allocation strategy for multicore real-time systems.  Their strategy exploits the relationship between the allocation of cache and memory bandwidth resources and a task's WCET to map tasks onto cores and to compute the resource allocation for each core, to fully utilize resources while ensuring timing guarantees.  Extensive evaluations using real-world benchmarks show that their strategy offers near optimal schedulability performance while being highly efficient, and that it substantially outperforms state-of-the-art solutions.

Linh, and her former/current doctoral students (Saeed Abedi, Neeraj Gandhi, Henri Maxime Demoulin, Yang Li, and Yang Wu), also won RTAS Best Student Paper Award.  Their paper titled “RTNF: Predictable Latency for Network Function Virtualization” presents a scalable framework for the online resource allocation and scheduling of NFV applications that provides predictable end-to-end latency guarantees.  RTNF is based on a novel time-aware abstraction algorithm that transforms complex NFV graphs and their performance requirements into sets of scheduling interfaces; these can then be used by the resource manager and the scheduler on each node to efficiently allocate resources and to schedule NFV requests at runtime.  Their evaluation, based on simulations and an experimental prototype, shows that RTNF can schedule DAG-based NFV applications with solid timing guarantees while incurring only a small overhead, and that it substantially outperforms state-of-the-art  techniques.

Congratulations to all!

Verisig is a tool for verifying properties of neural networks in autonomous systems.  The novelty of Verisig lies in its encoding of a deep neural networks as hybrid systems such that it can be easily composed with hybrid systems models of vehicle dynamics and verified using state-of-the-art solvers (e.g., Flow*).  Consequently, Verisig has been used to verify safety properties of learning-enabled closed-loop controllers containing neural networks with 10s of layers and 100s of neurons per layer.  Verisig 0.9 represents the first public release of the tool, being actively developed as part of the DARPA Assured Autonomy program.