2023 Cheriton Research Symposium

Welcome and Opening Remarks

Refreshments will be served

Blinded Memory

Outsourcing computing to a remote processor is popular and compelling. Cryptographic techniques like homomorphic encryption allow a client to outsource computation on sensitive data while ensuring that the data cannot be leaked. However, such techniques incur substantial computation and communication costs. Leveraging hardware assistance to efficiently ensure security is thus an attractive proposition. Trusted Execution Environments (TEEs), which saw widespread deployment in the early 2000s by mobile device manufacturers to run sensitive computations on commodity devices, can help to realize secure outsourced computing. But the security guarantees provided by traditional TEEs have been called into question by various recent attacks that exploit the inherent complexity of modern hardware and software.

In this talk, I will describe Blinded Memory (BliMe): on-going work by my students to design minimal processor extensions that can help to efficiently realize secure outsourced computing. BliMe consists of a minimal set of Instruction Set Architecture (ISA) extensions that use taint-tracking to ensure confidentiality of sensitive (client) data even in the presence of server malware, run-time attacks, or side-channel attacks. To secure outsourced computation, BliMe extensions can be used together with an attestable, fixed-function hardware security module (HSM) and an encryption engine that provides atomic decrypt-and-taint and encrypt-and-untaint operations. I will describe the overall architecture, the current status of the work, and the challenges we face.

Bio: N. Asokan holds a David R. Cheriton Chair and serves as the Executive Director of Waterloo Cybersecurity and Privacy Institute (CPI). Asokan’s primary research theme is systems security broadly, including topics like the development and use of novel platform security features, applying cryptographic techniques to design secure protocols for distributed systems, applying machine learning techniques to security/privacy problems, and understanding/addressing the security and privacy of machine learning applications themselves.

For more information about Asokan’s work, visit his website or follow him on Twitter/X.

The Future of Information Retrieval in the Era of Large Language Models

Information access — the challenge of connecting users to previously stored information that is relevant to their needs — dates back millennia. The technologies have changed — from clay tablets stacked in granaries to books on shelves arranged according to the Dewey Decimal Classification to digital content indexed by web search engines — but the aims have not. Large language models (LLMs) such as OpenAI’s ChatGPT and GPT-4, Google’s Bard and Gemini, Meta’s Llama 2, as well as numerous open-source models, represent the latest innovations that can help tackle this challenge. Nevertheless, the shortcomings of LLMs are well known, including hallucinations, out-of-date information, and inability to access private data.

In this talk, I’ll share my perspectives on the future of search in the era of LLMs. I argue that “traditional” retrieval tasks, far from being obsolete, are now more important than ever, as retrieval augmentation forms the basis of addressing exactly those challenges mentioned above. In this context, I’ll discuss how different techniques available today, such as sparse and dense retrieval models, contribute to the overall design of next-generation information access systems.

Bio: Professor Jimmy Lin holds the David R. Cheriton Chair in the David R. Cheriton School of Computer Science at the University of Waterloo. Lin received his PhD in Electrical Engineering and Computer Science from the Massachusetts Institute of Technology in 2004.

For a quarter of a century, Lin’s research has been driven by the quest to develop methods and build tools that connect users to relevant information. His work mostly lies at the intersection of information retrieval and natural language processing, with a focus on two fundamental challenges: those of understanding and scale. He is a Fellow of the ACM and a member of the SIGIR Academy.

The War Against Unpredictable Software

Software design and implementation is the greatest challenge that humans have ever faced. Done right, it is over a million times faster, cheaper and better than manually performing the same task. Done wrong, it can be a disaster. The last 50 years has created a mountain of software that is unpredictable, and more is being created every day. Yet, a key goal of engineering is predictability.

In this talk, I rail against unpredictable software (as a lonely warrior), particularly targeting parallel threads, and discuss a “solution” based on separate virtual address spaces, object-oriented programming and compiler-supported coroutine.

By invitation

Gelato, everyone is welcome!