Another week, another round.

---

Mandiant presents what AVDH (Agentic Vulnerability Discovery Harness) has been doing for them and how it fits into the existing AI landscape alongside CodeMender.

• https://cloud.google.com/blog/topics/threat-intelligence/staying-ahead-of-adversarial-ai-through-agentic-source-code-review/

---

Varonis reported a vulnerability to Microsoft in Copilot Personal. Almost as interesting as the one-click data exfiltration itself is the timeline. It was reported in December 2025, and the patches were released on 18.08.2026, the same day as the article. With the time from publication to exploit getting shorter and shorter, I find that a bit of a stretch.

…

Another week, another round. This week a little later and a tad shorter.

---

A Security has discovered multiple vulnerabilities in Zoom that allow a remote attacker in a meeting to hit all participants with a zero-click RCE.

• https://a.security/blog/asecurity-zoomsday

---

Talos has come across a new phishing framework. It works as live phishing, meaning the user directly enters their MFA for the attacker and essentially hands over access themselves.

• https://blog.talosintelligence.com/dissecting-the-jwr-phishing-framework/

---

Practical Security Analytics has shared their approach (tool-assisted) to adversary emulation. A good example of how red teaming doesn’t have to be purely about objective completion.

…

Another week another round:

---

A pretty interesting deep dive into browser fingerprinting from the Web Scraping Club.

• https://substack.thewebscraping.club/p/browser-fingerprinting-deep-dive

---

Cyble has published their ransomware threat report for the first half of 2026. Germany and France are leading the statistics on cases in Europe.

• https://cyble.com/blog/ransomware-threats-in-europe-h1-2026/

---

SpecterOps has published a four-part series on WSUS, including a Ludus lab and new tooling. An important prerequisite here is that the WSUS server uses an external SQL database that is not running on the same server.

…

Another week, another round.

---

Starting with a GitHub repo today. This is another intentionally vulnerable project, and having drones as the target makes for a nice change.

• https://github.com/nicholasaleks/Damn-Vulnerable-Drone

---

A colleague sent me an interesting article about hidden instructions in Word documents. These are later interpreted by CoPilot in Word and transferred into new documents.

• https://enklypesalt.com/posts/context-collapse-part3-ai-worming-through-word/

---

Another colleague sent me another article, covering an attack on EV charging infrastructure. There are significant security flaws in the charging stations, and according to the authors the vulnerabilities found can be applied to other manufacturers in a similar form. The blog has a variety of posts on charging infrastructure, so it’s worth having a browse.

…

Another week week, another round.

---

An interesting thought to design platforms to be safe not just for children but for people in general, with child protection added on top. The focus here is on doom scrolling, endless notifications, fake news, and so on. Things we consider unsuitable for children because they influence them also quite clearly influence adults.

• https://www.impactgrid.org/post/from-child-safe-to-human-safe-why-digital-platforms-need-universal-design-standards

---

Paradoxis found a CVE in Foxit PDF and published their writeup.

…

Another week, another round. There’s been a lot of AI content again this time, though it does make up a large portion of the news.

---

Elastic has analyzed a rapidly spreading new Malware-as-a-Service campaign. Teleputz relies on the increasingly popular ClickFix scheme for initial compromise. Reverse engineering is slowed down through garbage instructions in the code, and so far only a handful of C2 domains have been observed.

• https://www.elastic.co/security-labs/telepuz-maas-malware-clickfix

---

Mindgard has published a vulnerability in Codex (as of 16.07. it is still unpatched after a seven-month disclosure timeline). Codex searches for a Git executable in multiple locations within a project. If that executable is malicious, the consequences are fairly obvious.

…

Another week, another round. This time, the focus seems, by coincidence, to be more on escapes in various forms.

---

Synacktiv’s Paul Barbé presents a Helm template injection

• https://www.synacktiv.com/en/publications/charting-your-way-in-helm-template-injection.html

---

SpecterOps has released a new LLM jailbreaking tool

• https://specterops.io/blog/2026/06/29/llm-jailbreak-testing-with-jailbreaker/#h-experiments-when-one-prompt-becomes-a-matrix

---

The DFIR Report has analyzed and written up an Akira incident. A very detailed report, as usual, but also very exciting.

• https://thedfirreport.com/2026/06/29/from-bing-search-to-ransomware-bumblebee-and-adaptixc2-deliver-akira-3/

---

A short and crisp explanation of a container escape via IPv6, including a PoC

…

Another week, another round.

---

Socket looks at Polinrider’s approach, where Git history is deliberately manipulated and rewritten to the “main” branch via force push. This allows malicious commits to disappear among legitimate ones and the entire repo to appear OK.

• https://socket.dev/blog/polinrider-north-korea-linked-supply-chain-campaign-expands?utm_medium=feed

---

Stepsecurity maintains some GitHub actions. These were initially only available for paying customers, but are now also available for open source projects. It might be worth taking a look at their overview if you need a specific action.

…

Another week with articles. Thanks to the high temperatures I wasn’t motivated to do anything else but read, which lead to some more links this week.

---

Quarkslab took a closer look at Xiaomi’s proprietary security chip as used in cameras. Methods used included, among others, I2C sniffing, flash dumping, and firmware analysis.

• https://blog.quarkslab.com/black-box-probing-a-security-analysis-of-xiaomis-mja1-secure-chip.html

---

Praetorian’s setup for automated vulnerability analysis with AI was recently introduced. In the first part, the target was FreeBSD and the approach is explained in detail.

…

Another week, nother round

---

A VirtualBox escape through a use-after-free vulnerability, compiled by Luca Ginex.

• https://blog.exodusintel.com/2026/06/15/zombie-cotables-resurrecting-freed-memory-to-escape-virtualbox/

---

FIFA apparently was busy awarding peace prizes and didn’t have time to test their central systems. For many football fans, probably a happy outcome.

• https://bobdahacker.com/blog/fifa-hack

---

Niels Provos versus frontier models.

• https://www.provos.org/p/case-for-open-weight-models/

---

I have to agree with Justin Bollinger here. I’ve always taken a detour around jq and maybe used Python’s json.tool or awk and cut to prepare and filter data.

…

Another week with new articles.

---

Zoltan Madarassy and Alex Brown wrote another article for Elttam on what to watch out for in Go reviews.

• https://www.elttam.com/blog/golang-code-review-notes-ii

---

Matthew Green looked at the reasoning blocks in LLMs and was able to derive secrets from the systems (with a bit of luck). Additionally, neither Claude nor GPT5x seem to use system prompts in API mode.

• https://blog.cryptographyengineering.com/2026/05/29/fooling-around-with-encrypted-reasoning-blobs/

---

A comprehensive write-up of CVE-2026-23111. This is a use-after-free vulnerability caused by incorrect use of the exclamation mark.

…

Another week and another round of articles.

---

LevelBlue investigated a malware campaign in Brazil where Havoc was being used. Nine different stager variants were observed and they were distributed via .zip files. But that seems quite common, especially in May to receive invoices as ZIP archives. So good preparation.

• https://www.levelblue.com/blogs/spiderlabs-blog/the-demon-arrives-later-a-havoc-stager-hides-behind-microsoft-defender-dlp

---

Ruby Gems can now also be installed with a cooldown to reduce supply chain attacks. You can define the number of days a gem must exist before it’s listed for update.

…

Again a lot of interesting articles, at least in my opinion.

---

Interesting approach to bridge the air-gap using audio signals to get malware onto a target system when USB ports and the like are disabled. cocomelonc kicks off in the first blog post [1] with the transmitter side and some basics. The second part [2] then deals with the receiver under Linux.

1. https://cocomelonc.github.io/malware/2026/05/24/malware-tricks-56.html
2. https://cocomelonc.github.io/malware/2026/05/26/malware-tricks-57.html

---

Something a bit different for a change. nand2mario from Small Things Retro has set about pouring the 80386 (for those old enough to remember) into an FPGA based on the original microcode. Cool project.

…

Another week and a new round of links

---

Mathieu Farrell describes for Quarkslab which vulnerabilities he found in Optical Line Terminals in great detail. Including an explanation of what the whole thing actually is.

• https://blog.quarkslab.com/how-olts-may-have-exposed-entire-isp-networks.html

---

Massive attack on GitHub repositories. I couldn’t figure out which ones are directly affected. On 2026/05/21 it have been 5000 repositories and rising.

• https://www.ox.security/blog/megalodon-cicd-malware-github/

---

Elliot Belt/Felix Billières has extensively worked with AI in security research and provides his research along with recommendations on how to reproduce.

…

Another week, another round.

---

Datadog has looked at malicious agent skills and lists indicators to watch out for.

• https://securitylabs.datadoghq.com/articles/malicious-skills-supply-chain-risks-in-coding-agents-with-dynamic-context/

---

Synacktiv looked at the Tesla Wall Connector in 2025 and has now published a second part to the first article.

• https://www.synacktiv.com/en/publications/exploiting-the-tesla-wall-connector-from-its-charge-port-connector-part-2-bypassing.html

---

Trail of Bits introduces a new Go fuzzer.

• https://blog.trailofbits.com/2026/05/12/go-fuzzing-was-missing-half-the-toolkit.-we-forked-the-toolchain-to-fix-it./

---

Hacktron found an RCE in Github Copilot.

• https://www.hacktron.ai/blog/rce-in-vscode-copilot/

---

Daniel Stenberg explains “named globbing” in curl. Very practical and I just wasn’t aware that curl can do that.

…

Another Week, another round of interesting news. This time with different views on the Claude Mythos results from Mozilla and Curl

---

Citizenlab has compiled studies on telco network surveillance. Very comprehensive, but also very interesting.

• https://citizenlab.ca/research/uncovering-global-telecom-exploitation-by-covert-surveillance-actors/

---

Talos has identified a previously unknown attacker using the CloudZ RAT with a plugin “Pheno” that’s used to check if the Phone Link application is being used under Windows. The goal is probably to obtain credentials and possibly also OTP codes.

…

Another week, another round of interesting news.

---

Lazarus comes around the corner with a new malware variant for Macs and has ported a known variant from Go to Python using AI. The entry point is Telegram messages containing fake Teams, Meets, or Zoom invitations.

• https://any.run/cybersecurity-blog/lazarus-macos-malware-mach-o-man/

---

Talos sees strong growth in the adoption of Macs in the enterprise world. So far, Living off the Land tools seem to be less documented. They want to address this in their post. Certainly doesn’t hurt to be broadly positioned.

…

Another week another round.

---

Level Blue addresses a phishing vector that only became interesting thanks to the wider adoption of MFA. Attackers try to get the victim or helpdesk to deactivate or reset the second factor.

• https://www.levelblue.com/blogs/spiderlabs-blog/why-attackers-are-bypassing-phishing-emails-and-targeting-identity-instead

---

Calif unleashed Codex on a TV with a compromised browser to see how Codex performs in such scenarios and can turn an existing shell into a root shell. I think it’s a good insight into the current state of things and it aligns with what I’ve seen so far.

…

Another week another round

---

itm4n took a closer look at Windows’ implementation for collecting Bitlocker information. Maybe a pretty interesting topic if you want to dive into the depths of disk encryption.

• https://itm4n.github.io/bitlocker-little-secrets-the-undocumented-fve-api/

---

Something a bit different. Unlocking features in an Audi.

• https://blog.zsec.uk/audi-carplay-2026/

---

Something else again. A Mastodon discussion around the latest Windows zero-day Bluehammer. The discussion goes into the functionality a bit. Plus the repo.

• https://infosec.exchange/@wdormann/116358064691025711
• https://github.com/Nightmare-Eclipse/BlueHammer/tree/main

---

Talos analyzes LucidRook, a Lua-based stager that provides stripped Rust elements.

…

Another busy week:

---

Claude found an RCE in vim and emacs relatively easily.

• https://blog.calif.io/p/mad-bugs-vim-vs-emacs-vs-claude

---

Another week and another supply chain attack.

• https://www.stepsecurity.io/blog/axios-compromised-on-npm-malicious-versions-drop-remote-access-trojan
• https://securitylabs.datadoghq.com/articles/axios-npm-supply-chain-compromise/

---

And right along with that, TeamPCP uses audio steganography in WAV files to bypass EDR and other analysis tools.

• https://pwn.guide/free/cryptography/audio-steganography

---

I’m an avowed RSS fan, but I didn’t know about XSLT. Too bad I only find out about it when it’s being discontinued.

• https://justinjackson.ca/xslt

---

MDSec shows a way in the article to disable pre-boot DMA without triggering the Bitlocker Recovery Key by reading out the UEFI, modifying it, and writing it back.

…