Towards Reliable Binary Lifting

PhD Qualifying Examination


Title: "Towards Reliable Binary Lifting"

by

Mr. Xiang CHEN


Abstract:

Software binaries are the substrate of modern computing yet remain opaque to 
source- aware large language models (LLMs). Bridging this gap requires binary 
lifting: a multi-stage translation from raw bytes to high-level forms 
intelligible to humans and machines. As LLM-driven agents perform 
vulnerability triage, malware reverse engineering, and firmware audit on top 
of lifted output, lifter reliability has become the upstream bottleneck for 
every downstream task --- yet what reliability means here is contested.

This survey argues that reliability of binary lifting decomposes into three 
asymmetrically covered axes: operational equivalence, source fidelity 
(recovery of source-level constructs such as types, layouts, and symbolic 
data references), and input coverage. Across the four methodological 
paradigms developed for these axes — deductive, symbolic+SMT, empirical, and 
structural — the formal-semantics paradigms dominate operational equivalence 
but are silent on fidelity, while the measurement-based paradigms cover 
fidelity at the cost of universal operational guarantees. Four LLM-augmented 
hybrid paradigms compose these existing methods to close the cross-axis gap.


Date:                   Thursday, 28 May 2026

Time:                   10:00am - 11:00am

Venue:                  Room 2128A
                        Lift 19

Committee Members:      Prof. Charles Zhang (Supervisor)
                        Prof. Shing-Chi Cheung (Chairperson)
                        Dr. Shuai Wang