threatdeck
Live 14:45 UTC ·

New Technique Medium TD-2026-0018 For SOC / IR, AI Eng, Red Team AI-assisted

SentinelLABS releases VerifiedIDA, a standard and harness for LLM-driven reverse engineering in IDA Pro

SentinelLABS evaluated LLMs annotating IDA Pro databases and found inaccurate labels, unrecorded findings, lost unfinished work, and early stopping. The team proposed a standard requiring attributable evidence, verified database changes, preserved project state, and enforced review, and open-sourced VerifiedIDA as a reference implementation.

Why it matters

Defenders who rely on LLM-generated IDA annotations for incident response or detection work may act on labels that are wrong or incomplete. Attackers can use the same LLM capability to speed up vulnerability research and exploit development.

Do this

  1. Treat LLM-generated labels, comments, and types in disassembler databases as unverified until a reviewer checks them against the code they describe.
  2. Require that any LLM-assisted reverse engineering output record the evidence behind each annotation, including the binary and database revision it came from.
  3. Evaluate VerifiedIDA (open-source, IDA Pro) for workflows where LLM annotations feed detection development or threat hunting, and review its specification before adoption.
  4. Confirm that findings from LLM analysis are written to the database and not only kept in chat or session logs, and track any unfinished analysis outside the model's context.

Signals

  • NoCISA KEV
  • 25/100Priority
  • 80%Confidence

Evidence

“we developed and tested a standard for LLM-driven reverse engineering that keeps annotations tied to evidence, verifies database changes, preserves project state, and requires review of the model’s work.”

SentinelLABS

“We open-sourced VerifiedIDA , a reference implementation that applies the standard to autonomous reverse engineering in IDA Pro.”

SentinelLABS

“models produced inaccurate labels, left some findings unrecorded, lost track of unfinished work, and stopped the analysis early.”

SentinelLABS

Sources

Affected / related

LLM-driven reverse engineering, VerifiedIDA, IDA Pro, Ghidra, Binary Ninja, GPT-5.6 Sol, SentinelLABS

Published 2026-11-04 14:00 UTC · updated 2026-11-08 04:30 UTC · AI-assisted summary, checked against quoted sources · How we score

Following

Follow the vendors, products, AI tools, techniques and actors you care about. Matching stories, CVEs and exploited bugs are marked, a Following tab appears in the feed, and returning visits show what changed. Saved only in this browser.

Presets

Get updates

A personal RSS feed of new stories on everything you follow. Paste it into any feed reader, Slack or Teams RSS app.