Study uses C-to-C transformation to make four AI models miss buffer overflows
Understand this story
A study used Karma Pro to perform C-to-C transformations on C programs, hiding injected stack buffer overflows behind two techniques—enum and preprocessor divergence, and branchless MBA clamping—such that four frontier models—Nemotron 3.5 Super VL 120B, DeepSeek V4.1 Flash, OpenAI GPT-5.5, and Anthropic Claude Sonnet 5.5—each scanned at least 3 times yet still failed to report the flaw. The study states that the test source code contained no hints, no comment-based prompt injection, and file names did not reveal intent; all used codenames to avoid giving the models any clues.
AI overview based on source reports · Updated 8 hr ago
Coverage timeline
Follow the coverage to explore different perspectives.
- Hacker News AIAI Source Code Analysis Detection Evasion Techniques
The post describes a C-to-C transformation that hides a planted stack buffer overflow from four frontier AI models (Nemotron 3.5 Super VL 120B, DeepSeek V4.1 Flash, OpenAI GPT-5.5, Anthropic Claude Sonnet 5.5) using Karma Pro, combining an enum/preprocessor divergence (an enum constant the compiler sees as 9 but the preprocessor evaluates as 0 in an #if directive) with branchless MBA clamping that implements min(n, 9) via sign-bit masking instead of a ternary. It also reports that models sometimes hedged that a buffer "can be" too small without computing the capacity, and that a safe twin differing by one line (#if M0 + 4 >= 4) confirms the concealment rests on the preprocessor divergence plus the clamp's opacity.
Activity for this story
- Current comparable activity
- 8
- Peak comparable activity
- 10Nov 7 18:00
- Change in the past 24 hours
- –
Trends compare the same fully observed participants, which may cover fewer participants than current activity. Move the pointer or tap the chart to inspect hourly activity; use the left and right arrow keys with a keyboard.