QueryX: Symbolic Query on Decompiled Code for Finding Bugs in COTS Binaries
Why this work is in the frame
A frame that forgets how it found something cannot be audited. These are the routes that admitted this work.
Bibliographic record
Abstract
Extensible static checking tools, such as Sys and CodeQL, have successfully discovered bugs in source code. These tools allow analysts to write application-specific rules, referred to as queries. These queries can leverage the domain knowledge of analysts, thereby making the analysis more accurate and scalable. However, the majority of these tools are inapplicable to binary-only analysis. One exception, joern, translates a binary code into decompiled code and feeds the decompiled code into an ordinary C code analyzer. However, this approach is not sufficiently precise for symbolic analysis, as it overlooks the unique characteristics of decompiled code. While binary analysis platforms, such as angr, support symbolic analysis, analysts must understand their intermediate representations (IRs) although they are mostly working with decompiled code.In this paper, we propose a precise and scalable symbolic analysis called fearless symbolic analysis that uses intuitive queries for binary code and implement this in QueryX. To make the query intuitive, QueryX enables analysts to write queries on top of decompiled code instead of IRs. In particular, QueryX supports callbacks on decompiled code, using which analysts can control symbolic analysis to discover bugs in the code. For precise analysis, we lift decompiled code into our IR named DNR and perform symbolic analysis on DNR while considering the characteristics of the decompiled code. Notably, DNR is only used internally such that it allows analysts to write queries regardless of using DNR. For scalability, QueryX automatically reduces control-flow graphs using callbacks and ordering dependencies between callbacks that are specified in the queries. We applied QueryX to the Windows kernel, the Windows system service, and an automotive binary. As a result, we found 15 unique bugs including 10 CVEs and earned $180,000 from the Microsoft bug bounty program.
Fetched live from OpenAlex and de-inverted. Abstracts are not stored in this database: the inverted indexes are 8.6 GB of the frame’s 9.3 GB of text, and the host has 13 GB free.
Full frame distilled prediction
Teacher imitationNot calibrated prevalence, not ground truth. Human validation pending. Learned from the 10,348 direct Codex labels and 10,348 direct Gemma labels. Candidate is the union of thresholded teacher heads; consensus is their intersection. These outputs are machine_predicted_unvalidated and are not human labels or direct frontier model labels.
Codex and Gemma teacher scores by category
| Category | Codex | Gemma |
|---|---|---|
| Metaresearch | 0.001 | 0.001 |
| Meta-epidemiology (narrow) | 0.000 | 0.000 |
| Meta-epidemiology (broad) | 0.000 | 0.000 |
| Bibliometrics | 0.000 | 0.001 |
| Science and technology studies | 0.000 | 0.000 |
| Scholarly communication | 0.000 | 0.000 |
| Open science | 0.001 | 0.000 |
| Research integrity | 0.000 | 0.000 |
| Insufficient payload (model declined to judge) | 0.000 | 0.000 |
Machine scores (provisional)
The two teacher heads of the student model, read on this work. A score orders the frame for review; it never asserts a category, and the validation status ships verbatim with every row.
Baseline scores from an immature model (maturity gate not passed, 7 training rounds). Scores rank; they never assert a category.
score_only:v0-immature-baseline · verbatim from the scoring run: score_only means the number may rank works, and no category label ships from it