Paper Image

Language models for vulnerability detection

Published on:

7 November 2023

Primary Category:

Machine Learning

Paper Authors:

Benjamin Steenhoek,

Md Mahbubur Rahman,

Shaila Sharmin,

Wei Le

Bullets

Key Details

Analyzes if language models learn code semantics for vulnerability detection

Compares model interpretations, attention, interaction matrices to bug semantics

Better models align more to vulnerability statements

Adding annotations of vulnerability semantics improves models

AI generated summary

Language models for vulnerability detection

This paper analyzes whether pretrained language models can learn semantics of code relevant for vulnerability detection. The authors compare model interpretations, attention, and interaction matrices to buggy paths and potentially vulnerable statements. They find better models align more to vulnerability semantics, but alignment is still low. Adding annotations of vulnerability semantics improves model performance and alignment.

Answers from this paper

Comments

No comments yet, be the first to start the conversation...

Sign up to comment on this paper

Sign Up