Paper

DIG: Oracle-Guided Directed Input Generation for One-Day Vulnerabilities

Unreadunread

AI summary

arXiv:2606. 13037v1 Announce Type: new Abstract: One-day vulnerabilities pose significant risks due to delayed or incomplete patch adoption.

Reading status

Priority

About this paper

Authors: Andrew Bao (University of Minnesota, Twin Cities), Haochen Zeng (University of California, Riverside), Peng Chen (Independent Researcher), Stephen McCamant (University of Minnesota, Twin Cities), Pen-Chung Yew (University of Minnesota, Twin Cities)
Released: 2026-06-12 (2mo ago)
Topic: Related to current project
Relation to my work: Read next because DIG: Oracle-Guided Directed Input Generation for One-Day Vulnerabilities overlaps with clean result "LoRA persona trained on <A> alone emits <B> at 23.5% when a co-trained partner learns <A>...<B>, vs 0% control on Qwen2.5-7B-Instruct (MODERATE confidence)", clean result "Leakage rate is a usable signal for recovering trigger-shaped phrases on Gaperon-1125-1B without knowing the hidden trigger itself (MODERATE confidence)", clean result "Language-mismatch LoRA SFT on Qwen2.5-7B leaks the trained completion language into bystander directives the model was never trained on, absent under same-language SFT (LOW confidence)". Matching terms: code, strong, rect, eval, line, compare. Source: arxiv cs.CR (Cryptography and Security).
arXiv: 2606.13037

Ask Claude about this paper

Posts a comment with @claude so the runner spins up a Claude Code session scoped to DIG: Oracle-Guided Directed Input Generation for One-Day Vulnerabilities. Replies appear in the conversation below.

Comments

No comments yet.

Related records

No links yet.