Loading story…
Lambda Security introduced Frame, a neuro-symbolic static application security testing (SAST) system that pairs a sound separation-logic analysis core with a verified, tiered large-language-model layer. The design aims to ground LLM-assisted vulnerability findings in symbolic reasoning while extending analysis to code patterns the core engine cannot resolve alone.
On Endor Labs’ real-world code corpus, Frame reportedly achieved 0.67 recall and 0.51 precision, compared with Semgrep’s reported 0.52 recall and 0.40 precision. The project also reported finding about 65 vulnerabilities across Java, JavaScript/TypeScript, and C# that were missed by both Frame’s symbolic engine and Semgrep.
Track how attackers are adapting to this technology.
1 event from the most recent confirmed update back to the earliest known activity.
The neuro-symbolic SAST system Frame was evaluated on Endor Labs' real-world corpus, reporting 0.67 recall and 0.51 precision. In the stated comparison, it identified approximately 65 Java, JavaScript/TypeScript, and C# vulnerabilities missed by both its symbolic engine and Semgrep.
Follow how adversaries are adapting to this technology, and where it touches your stack today.
2 references tracked. Mallory keeps watching after this page renders.
reddit.com
Open sourcelambdasec.github.io
Open sourceMap indicators from this story to your assets and identify affected systems in minutes.
Every observed campaign, victim, and pivot linked to actors named in this story.
Malware, exploits, and IOCs connected to the activity described here.
YARA, Sigma, and Snort rules deployed to your SIEM as soon as they’re published.
Get matching new stories delivered to your team as they break — not the next morning.
Ask questions about this story and take action on the answers.
Other stories Mallory is tracking that overlap on entities, sources, or topic labels.
SentinelOne detailed a multi-agent LLM pipeline for automated static malware analysis that uses an orchestrator and tool-specific subagents for **radare2**, **Ghidra**, **Binary Ninja**, and **IDA Pro** to examine macOS samples in sequence. The system, built on OpenClaw, keeps shared context in memory, normalizes outputs into a common schema, and routes findings to a report writer only after a second adversarial review stage known as **the Gauntlet**. The design favors deterministic bridge scripts over MCP-style integrations to improve repeatability, reduce latency, and support batch analysis at lower cost. Testing on malware including **SysJoker**, **WizardUpdate**, **FinderRAT**, and a Go-based infostealer showed the pipeline could catch incorrect C2 paths, false Tor indicators, malformed strings, dead code, and decompiler-induced hallucinations that might otherwise appear in analyst reports. Related writeups and community discussion echoed the same conclusion: single-tool LLM malware analysis can produce authoritative-looking but unreliable results, while a serial consensus approach that anchors claims to virtual addresses and decompilation snippets can materially improve confidence in automated reverse-engineering output.
Jun 29, 2026
Security researchers highlighted multiple ways **LLM adoption can introduce or amplify risk**, including both technical attacks and unsafe development practices. G DATA reported that a Git-hosted “detector” for the **Shai-Hulud worm** shipped with “test files” that were effectively *real malware*: scripts capable of deleting user directories and, in at least one case, uploading data to actual threat actors. The files were apparently intended to validate detection efficacy and may have been produced via AI-assisted “vibe coding,” where the model replicated malicious behavior one-to-one while comments claimed the code was only a simulation; although the test artifacts are not executed during normal tool operation, users could trigger damage by manually running them. Separate academic work summarized by Bruce Schneier described **side-channel attacks against LLM inference**, where data-dependent timing and token/packet-size patterns (including those introduced by efficiency techniques like speculative decoding) can leak information about user prompts even over encrypted channels. Reported impacts include inferring conversation topics with high accuracy and, in some settings, recovering sensitive data such as phone numbers or credit card numbers via active probing. In parallel, an SC Media segment discussed the operational upside of **LLM-driven secure code analysis**, citing results that improved security across hundreds of open-source projects but noting the importance of human validation and patching effort; an OSINT Team post provided a cautionary, practitioner-level example of how easily malware can be accidentally executed during analysis, reinforcing the need for disciplined handling and isolation when working with suspicious files.
Jun 29, 2026
Security researchers and vendors reported that large language models are becoming materially useful across offensive and defensive security workflows, including vulnerability discovery, exploit development, autonomous penetration testing, and cyber threat intelligence extraction. Recent work described LLM-assisted systems that generated proof-of-concept exploits for npm flaws at a reported **77%** success rate, exploited real-world web vulnerabilities with multi-agent architectures, and executed multistage attacks in emulated enterprise networks when paired with task-specific agents and attack-abstraction layers. Other experiments found local self-hosted models could reliably solve straightforward Juice Shop challenges, while purpose-built scaffolds helped researchers uncover memory-corruption bugs in Windows endpoint products and accelerate reverse engineering of AV and EDR logic. At the same time, multiple studies warned that headline results can overstate real-world capability without strong validation. Google Project Zero said benchmark design, tooling, and automatic verification heavily influence measured performance and concluded current models still fall short of meaningful autonomous offensive research in live environments. Separate academic and practitioner reviews found many automated pentesting frameworks remain brittle, and human validation showed that **71.5%** of supposedly successful LLM-generated exploit PoCs in one follow-up study were actually invalid because the models simulated exploitation rather than triggering the flaw. Across the research, the consistent finding was that LLMs perform best when constrained by structured workflows, specialized tools, execution feedback, and rigorous verification rather than being trusted as fully autonomous hackers.
Jun 29, 2026