<i>CPA</i>: Accurate <i>C</i>ross-<i>P</i>latform Binary <i>A</i>uthorship Characterization Using LDA
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
Binary authorship characterization refers to the process of identifying stylistic characteristics that are related to the author of an anonymous binary code. The aim is to automate the laborious and error-prone reverse engineering task of discovering information related to the author(s) of binary code. This paper presents CPA, a novel approach for characterizing the authors of program binaries. Instead of using generic features such as n-grams, CPA proposes a set of new features based on collections of various aspects of author style, including author code traits, code structure characteristics, and author expertise in solving coding tasks. It employs the Latent Dirichlet Allocation (LDA) algorithm to generate author style signatures to help identify similar author style characteristics in other binaries. We evaluated CPA on large datasets extracted from selected opensource C/C++ projects in GitHub and Google Code Jam events, and it successfully attributed a large number of authors with a significantly higher F <sub xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">1</sub> score: around 91% when the number of authors was 1,500. In addition, the false positive rate was low, around 1.5%. When the code was subjected to refactoring techniques or code transformation or was processed using different compilers/compilation settings, there was no significant drop in accuracy, demonstrating the robustness of our tool. Finally, in the case of code written by multiple authors, CPA was able to identify the authors with a high F <sub xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">1</sub> score, around 89%.
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 | 0.000 | 0.000 |
| 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 | 0.004 |
| Open science | 0.000 | 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