Kernel Extensibility with eBPF
eBPF is an exciting Linux-supported framework that allows userspace applications to write their custom functions that can be run in the kernel. We are exploring new use cases for eBPF, such as accelerating storage and database requests, allowing users to tweak and modify different kernel policies, and exploring how eBPF can be used as an interface to program various hardware accelerators.
Recent work: XRP [OSDI'22], BPF-oF [Arxiv'23]
Recent work: XRP [OSDI'22], BPF-oF [Arxiv'23]
Software Support for Tiered Memory
We are seeing the rise of new and exciting memory and storage technologies, such as CXL, non-volatile memory and low-latency and dense flash. These technologies have the potential to change fundamental assumptions in how software systems are built across all the layers of the stack, from the OS to distributed storage systems and databases. Our research focuses on how to take advantage of these new memory technologies to build fast, reliable and scalable software systems.
Recent work: CXL [OSDI'24], PrismDB [ASPLOS'23]
Recent work: CXL [OSDI'24], PrismDB [ASPLOS'23]
Distributed Databases Leveraging Fast Datacenter Networks
With the wide availability of low latency networking and storage, classical distributed database protocols, such as 2PC, are no longer a scalability bottleneck when building distributed transactional systems. Armed with this observation, we are designing the next generation of distributed transactional databases that are both fast (millions of transactions per second) and general (full SQL, strong semantics).
Recent work: Chablis [CIDR'24], Chardonnay [OSDI'23]
Recent work: Chablis [CIDR'24], Chardonnay [OSDI'23]
Differentially-Private Systems
|
Differential privacy is a technique where by adding noise to computation we can provide strong guarantees on the amount of data that is leaked from a database or ML model. An increasing number of applications are turning to differential privacy to protect user data as well as comply with privacy regulations. However, there are many practical obstacles in adopting differential privacy in real-world systems, including how to schedule, budget, cache and manage the differential privacy budget. In our research, we explore many of these problems, with the vision of democratizing privacy.
Recent work: Turbo [SOSP'23], PrivateKube [OSDI'21] |
|
|
Data-driven Cyber Fraud DetectionAttackers are increasingly tailoring their cyber attacks to an individual target, and using social engineering to fool their victims to click on a phishing link or open a malicious file. Our research characterizes these targeted attacks and use ML to build automated detectors that can automatically stop them that can be deployed in real-world systems.
Recent work: BECGuard [UseSec'19], Lateral [UseSec'19] |