Weaknesses in LLM-Generated Code for Embedded Systems Networking
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
Modern firmware development is done in a fast-paced, time-constrained environment. This pressure tempts developers to use generative AI to write code for them to save time. While this is a powerful tool with careful developer review, these reviews are commonly sacrificed to meet deadlines. This results in AI-written code existing verbatim, deployed in the firmware of devices finding their way into our cyber-physical environment. In the absence of developer oversight, we suggest that generative AI-written code does not sufficiently account for common software weaknesses. In this work, we explore a collection of modern Large Language Models (LLMs) and use them to generate code based on popular network standards. We fuzz this code to discover vulnerabilities in the code generated by the LLMs. We organize these vulnerabilities according to the Common Weakness Enumeration (CWE) and use them to develop a three-axis taxonomy of common LLM-generated weaknesses. Finally, we provide suggested input categories to more easily exploit these weaknesses in a black-box setting, as a first step towards fuzz testing for LLM-generated code in embedded systems networking.
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 | 0.002 | 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 | 0.001 | 0.001 |
| 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