Iru Threat Intelligence (@iruthreatintel) header
Iru Threat Intelligence (@iruthreatintel) profile picture

@officiallyiru Security Research team's space for sharing intelligence and observations about what we see in the technology and cyber industries.

Joined September 2026
Photo by Iru Threat Intelligence (@iruthreatintel)Photo by Iru Threat Intelligence (@iruthreatintel)Photo by Iru Threat Intelligence (@iruthreatintel)Photo by Iru Threat Intelligence (@iruthreatintel)Photo by Iru Threat Intelligence (@iruthreatintel)
Rockstar’s big launch is today. We've got one on Dec 9. Fewer carjackings, but it might lower your org’s compliance wanted level. ⭐⭐⭐⭐⭐
Speaking of big launch events, save the date for December 9th to hear some very exciting updates from us!
5
218
Talks day 2! Don't miss @officiallyiru's own @theevilbit at #OBTS with "macOS Exploit Mixtape: Hack Like It's the 80s" youtube.com/watch?v=PBYs9omW…
5
15
1,011
Google’s GTIG documented an AI framework that diffs a patch, finds the flaw it fixed, and builds an exploit without human input. Patch lag is the exposure. GTIG also names frameworks in threat actor use: ARTEX, CAIRN, CyberStrikeAI, PentAGI. We’re testing them... which first?
BREAKING: Google says “the era of autonomous attacks has arrived.” Hackers are now using AI to scan software updates, figure out which security flaw each patch fixes and build an attack before companies can install the fix, all without a human involved, according to a new report from the Google Threat Intelligence Group. “Defending against machine-speed threats at human speed isn’t feasible,” Google said.
3
212
Day 2 of trainings at #OBTS. Couldn't ask for a better place to learn from the best in macOS security.
3
8
1,014
First day of Trainings at #OBTS !
4
2
28
2,645
Iru Threat Intelligence retweeted
Enabling safe, productive AI usage is the job. Not blocking it. Iru AI Governance controls access across every AI app based on the groups you define. Web and desktop can differ for the same app, so one tool can be allowed in the browser and blocked as a desktop app.
1
1
7
5,103
Iru Threat Intelligence retweeted
Wheels up for #OBTS! ✈️🌴 The (stylish) #OBTS crew is trading gray skies for Hawaiian sunshine as we head to the 9th edition of Objective by the Sea! Talks, trainings, networking & waves ...see you by the sea! 🤙
🤖 Made with AI
6
26
1,381
Iru Threat Intelligence retweeted
Ok another macOS stealer/backdoor. (artifacts look similar to previous reports) Starting with a compiled (0 VT hits) applescript file: GoogleHelper.scpt: 77f9762cdf9d26c0a1fc9c7a5353091093ff9d21ce46a9609004791094626372 There are 3 stages I'll summarize the best I can. :) 🧵
1
9
2
42
3,120
Aloha, #OBTS! 🌺 Next week, part of the Iru team will be in Maui for the best Apple security conference around, and we can’t wait to catch up with everyone there. Come say hi! Shoutout to our own Csaba Fitzl (@theevilbit), who’s speaking with @gergely_kalman on hacking macOS like it’s the 80s 🎸 Thanks to @objective_see for hosting! 🍎
6
2
13
3,167
Any good incident response plan should call for a deep-dive analysis for lessons learned. At minimum, teams should be asking: “What happened, and how do we do better moving forward?” Even if the event doesn’t happen to your organization, you have an opportunity to learn the lesson without paying the price. Using Binance as an example, the report describes social engineering through a fake recruiter, a tampered Homebrew install, credential theft, lateral movement using an internal AI tool, and abuse of legitimate deployment permissions. The human element is the obvious issue, but there’s a question that hasn’t gotten much attention: would the later attack have happened without the June tampered Homebrew install? That incident wasn’t what ultimately led to the theft, but Binance says the engineer and his wallet-team colleagues remained exposed afterward. The initial event may have given the attackers early insight into who to target next. We looked into Homebrew in 2025, around the time of the Shai-Hulud npm worm, and found spoofed installer sites targeting developers. To be clear, what we describe is not the same technique, but a similar class of threat: iru.com/blog/brewing-trouble… Attacks like this could become more common as shared dependencies and AI-augmented coding become a bigger part of how software gets built. We are of the opinion that security teams need to treat not only the packages they install as sources of risk, but also the package managers themselves. nitter.cf/binance/status/2117626… #IncidentResponse #Homebrew #ThreatIntelligence
1
3
370
Disclosures are up ~2.3×. CISA KEV additions are up about 31%. The share of new CVEs that reach confirmed exploitation has fallen. So has the actual risk scaled with the volume, or are we mostly looking at a bigger haystack? If volume is a poor proxy for risk, how should security teams adjust as discovery accelerates? Inspired by the “vulnerability tsunami” and “Black October” framing, we stepped back to look at the bigger picture and what practitioners should actually prioritize. Full analysis found here: iru.com/blog/black-october #VulnerabilityManagement #ThreatIntel #blackoctober
1
2
9
460
If you've been anywhere near security social media in October, you've heard the language. "Black October." A "vulnerability tsunami." Microsoft commented that AI is changing the physics of cybersecurity. We even saw a "vulnpocalypse" thrown out there. The rhetoric isn't empty: disclosures have surged, including bugs in some of the most reviewed code in the world. However, volume itself doesn't necessarily translate to risk. This made us wonder - what important trends can we see if we demystify the hype and look past the raw volume? How does AI fit in this picture? What does this really mean for security practitioners and how should teams adjust their strategy? Our full analysis on these questions drops next week.
1
141
Iru Threat Intelligence retweeted
有人发现了一种非常离谱的 JavaScript 混淆技术:通过在代码里塞入大量生物武器、毒药和致死剂量相关的函数名,让 AI 因为安全限制而拒绝分析代码。 安全研究员 veritas 在分析浏览器指纹脚本时,发现一段来自 ServicePipe 相关域名的 JavaScript,里面充斥着极其诡异的函数名: computeRicinLethalDosageForWaterSupply calculateDeadlyLethalSuicideAspirinDosage 光看名字,仿佛在研究生化武器。但实际上,这段代码只是在收集浏览器指纹,包括 User-Agent、GPU、Canvas 等信息。 研究员怀疑,这些名字是开发者故意设置的陷阱,专门用来阻止 AI 辅助逆向工程。 于是他做了个实验: • Astra:拒绝分析 • Astra + Daybreak Blue:仍然拒绝 • GLM-5.3:成功分析 这就有意思了。 传统代码混淆是通过复杂控制流、加密字符串、无意义变量名,让人类和工具难以理解代码。 而这种新思路,直接瞄准了大模型的安全机制。 代码本身可能完全无害,但只要在函数名里塞进足够多的危险词汇,部分 AI 就可能主动放弃分析。 更讽刺的是,模型越严格遵守某些安全规则,面对这种混淆时反而可能越容易受到干扰。 当然,目前还没有证据证明开发者一定是故意这样设计,也不能说明这种方法对所有模型都有效。 但它揭示了一个值得关注的趋势:随着越来越多的逆向工程依赖 AI,未来的代码混淆技术可能不再只针对编译器、反编译器和人类研究员,还会专门针对大模型的安全判断机制。 以前是让代码难以理解,现在甚至可以尝试让 AI 不敢理解。
236
1,381
462
9,660
1,135,195
Iru Threat Intelligence retweeted
Last year our audits ran by hand on spreadsheets, plus a temp contractor, after our old compliance tool's integrations failed us. This year with Iru Compliance Automation we spent about 35% less time on the audit without a contractor. Much of that came from evidence that collects itself and gets mapped once across four frameworks.
1
2
2
6
30,131
Iru Threat Intelligence retweeted
I got assigned 5 new 🍎 CVEs. Update your boxes. apple-cve.com/ also contains the latest advisory updates.
5
7
2
99
5,532