Derek Banks (@0xderuke) header
Derek Banks (@0xderuke) profile picture

AI and Data Nerd. Black Hills Infosec

VA
Joined August 2010
Photo by Derek Banks (@0xderuke)Photo by Derek Banks (@0xderuke)Photo by Derek Banks (@0xderuke)Photo by Derek Banks (@0xderuke)Photo by Derek Banks (@0xderuke)Photo by Derek Banks (@0xderuke)Photo by Derek Banks (@0xderuke)Photo by Derek Banks (@0xderuke)Photo by Derek Banks (@0xderuke)Photo by Derek Banks (@0xderuke)
Derek Banks retweeted
We caught another threat actor using zero-day exploits to compromise an edge device — in this case Ivanti Connect Secure. The used a few different exploits to facilitate RCE and then an escalation of privilege. They did some interesting anti-forensics tricks and then largely leveraged tooling they brought over to try and move laterally. We were able to fully recreate the exploit chain from memory and core dumps! We built a custom analysis profile and analyzed the memory in @Volexity's Volcano software. Details of the exploits, the threat actor, and how we analyzed it are all up on our blog!
.@Volexity has discovered in-the-wild exploitation of a #zeroday exploit chain targeting Ivanti Connect Secure VPN appliances (CVE-2026-104935). Beginning October 3, a previously unknown threat actor, UTA0576, chained three vulnerabilities to gain unauthenticated root-level code execution on a fully patched device. Notably, UTA0576 deployed no malware or web shell. Instead, the threat actor re-exploited the device for every command it ran and tampered with logs to cover its tracks, evading Ivanti’s Integrity Checker Tool. Using its #Volcano memory analysis platform, Volexity analyzed the compromised device’s memory to uncover the full exploit chain and the attacker’s post-exploitation activity, which focused on credential theft, Active Directory reconnaissance, and data exfiltration. Volexity reported these findings to Ivanti through coordinated disclosure. Read the full analysis, including detection guidance and IOCs, in our blog post here: volexity.com/blog/2026/10/13… #DFIR #threatintel #memoryforensics
2
20
3
68
6,020
Derek Banks retweeted
Replying to @mcuban @grok
My area here… No, we have not seen this yet. AI is really good at speeding up existing attacks, reverse engineering, payload development, and exploit dev. AI is fast, but kinda dumb. cloudsecurityalliance.org/ar…
1
1
11
388
Derek Banks retweeted
Apparently all of these “AI is escaping containment” stories are actually just AI researchers not knowing jack squat about the absolutely most basic security practices.
265
910
96
9,538
186,456
Derek Banks retweeted
An agent is not a user, and it is not a service account. It is closer to a brilliant intern with infinite energy, no instinct for boundaries and whatever credentials you handed it. A swarm is just that intern, copied a thousand times. Claims of an unstoppable internet takeover skip what every swarm still runs on: tokens, compute, credentials, infrastructure, and access. Every one of those dependencies has a cybersecurity control point on it. Identity management. Least privilege. Segmentation. Work done by people who write the security playbooks and have spent decades securing networks for a living. The swarm debate has been run as an AI safety conversation. The people who would actually pull these security levers and the leaders of this field need seats at the table in briefings, on committees, wherever policy is being written.
4
4
2
28
2,997
Derek Banks retweeted
Just for your information: We (offensive or defensive cybersec people) and threat actors don't have the same capabilities as the frontier labs do. The API cost of testing ExploitBench like they did would be a whopping ~$59.3M (Astra). Oh, and you cant do it like they do, they remove all guardrails and classifiers. What about renting AI hardware in the cloud? DeepSeek v4? $3.4M Kimi K3? $16.9M GLM 5.3? $9.3M Not to mention the engineering cost to make a harness operate 10,000 agents. (Napkin-Math based on the Huggingface reports. Exploit bench eval consisted of 10,000 concurrent agents, 8day non-stop run, no limiters)
43
104
10
829
57,105
Derek Banks retweeted
If you're worried about two AIs communicating by heating CPUs, please tell me what sort of bandwidth in bits per minute you expect to be able to extract from it. In capital letters: PLEASE DO NOT ASSUME HYPERINTELLIGENT AI IS NOT SUBJECT TO PHYSICS.
63
66
10
531
14,743
Derek Banks retweeted
Again more doomer bullshit. And I quote “and this is mostly academic” Yes Noam, it is highly academic. What you’re referring to is most likely BitWhisper and/or perhaps or thermal covert channels on multi-core platforms or maybe even HVACKer which is a more recent extension of the work detailed in BitWhisper. Since you’re fear mongering like mad, perhaps someone should explain to the good people how ridiculous your statement is. Let’s start with BitWhisper. BitWhisper was published in the 2015 IEEE Computer Security Foundations Symposium (CSF). In this paper they experimentally tested machines positioned approximately 0–40 cm apart and reported effective communication rates of approximately 1–8 bits per hour. 1-8 bits per hour at a max of 40cm away! Now that’s a close distance for any true air gap and extraordinarily slow to take over the world don’t you think? Sure it may work for short commands or small secrets. But the statement you’re making is ludicrous. HOWEVER, any serious security professional understands that part of air gapping a machine should include shielding. You know like RF/EM shielding, acoustic shielding, optical shielding, USB/peripheral controls, power-line filtering/isolation, and yes thermal shielding, in addition to physical separation. Now what about thermal covert channels on multi-core platforms? That’s a goodie from USENIX Security 2015. Instead of transmitting heat between separate computers, they demonstrated communication between supposedly isolated components inside the same machine. Their experimental Intel Xeon system achieved thermal covert-channel rates of up to 12.5 bits/second. Now to your credit, a later review of covert-channel research explicitly categorizes temperature-based channels as working within a processor core, between processor cores, and between adjacent desktop computers. But again I point you back to what a real air gap is. And how it addresses, as in eliminates, the remote possibility of what you’re stating in the interview. So, on behalf of humanity we’re not mushrooms my man. So stop keeping us in the dark and feeding us bullshit. But hey, I get it. If any of you doomers set up a proper test environment you’d disprove all of your own assertions. 😂 Cc @mattjay @RSnake For reference: “BitWhisper: Covert Signaling Channel between Air-Gapped Computers Using Thermal Manipulations” by Mordechai Guri, Matan Monitz, Yisroel Mirsky & Yuval Elovici published March 26, 2015 (arXiv); subsequently published at IEEE CSF in July 2015. arxiv.org/abs/1503.07919 “Thermal Covert Channels on Multi-core Platforms” by Ramya Jayaram Masti, Devendra Rai, Aanjhan Ranganathan, Christian Müller, Lothar Thiele & Srdjan Capkun published March 24, 2015 (arXiv); published at USENIX Security in August 2015. ⁠arxiv.org/abs/1503.07000 “HVACKer: Bridging the Air-Gap by Attacking the Air Conditioning System” by Yisroel Mirsky, Mordechai Guri & Yuval Elovici published March 30, 2017. arxiv.org/abs/1703.10454
OpenAI's Noam Brown says air-gapping the computers may not stop a misaligned AI, because two air-gapped machines can still talk by running a CPU hot and reading the temperature change "But I think the major takeaway from the incident is that people underestimated the AI. And we never want to be in a situation again where we underestimate the AI. It's a weird world, because AI progress is so fast that people are consistently underestimating the AI." "So to be in a situation where you don't underestimate it again, when it comes to safety and alignment, you have to have a very, very, very high bar." "You could even go as far as to say, "Well, we should air gap the computers." And I'm not convinced that that would be sufficient." "There are studies, and this is mostly academic, where you can have two computers next to each other that are air-gapped and they're still able to communicate with each other because they have temperature sensors." "One of them is able to run their CPU really hot, and then the other one can actually detect the temperature change, and then that actually gives them a mechanism to communicate." _________ Link and more key quotes from OpenAI's safety related conversations: firesidealpha.substack.com/p…
35
61
22
302
24,660
Derek Banks retweeted
One of the biggest issues facing infoec isn't AI. It's institutional memory loss. We forget how we got here, how most breaches actually happened over the years, how vendors historically treated security, and which controls repeatedly failed. Then every new technology shows up and we debate it like none of that history exists.
26
39
3
202
10,219
Sounds good. Let’s start here: Don’t ask Cybersecurity firms for help and then ignore us when we apply for access to your cyber models.
this is a critically important moment for cyber defense with AI; there is not much time to act. we are happy if you want to work with us or any of our competitors or partners, but please take this moment seriously. only an urgent and intense collective response will work.
9
9
1
93
5,067
I always look forward to when we, CISA’s red team, gets to publish our reports, and today is one of those days! “A Tale of Two SOCs” - a story where we targeted two different orgs, with the same tradecraft, and very different responses. Read it! - cisa.gov/news-events/cyberse…
4
75
7
294
45,428
It’s really unfortunate to see this from one of my favorite thinkers. China has been hacking the entire planet and pillaging its intellectual property for decades. It is extraordinarily naive to think they are handing out open models because they care about the world. They are doing it because they believe it is the best method for undermining US AI labs and being first to achieve AGI and ASI. If you don’t think they would like the entire world to be a lot like Hong Kong, then you are not paying attention.
Frontier Labs are incented to be fearful as it provides a moral justification for staying closed. Chinese Labs take the consistent position of doing what’s good for all and being open. So the little guys find themselves in the uncomfortable position of rooting for the Chinese…
22
8
2
112
21,412
Derek Banks retweeted
Want to stop npm supply chain issues? Make three changes. 1. Use pnmp 2. Disable script execution 3. Use renovatebot but when a pin happens, delay it 3 days. Most are detected pretty quickly and within hours, the delay 3 days for updates gives a nice buffer on if a compromised package is there. Script execution should eliminate anything I've seen on compromised npm packages period.
11
33
3
209
13,612
Derek Banks retweeted
When discussing AI safety, it's critical to understand: 1. Models can only act on the world when we give them a path with which to do so 2. Everyone understands prompts aren't guardrails 3. Anyone discussing a real-world safety issue must admit it flowed from a lack of their own technical controls
10
21
1
85
7,089
Derek Banks retweeted
I’m generally opposed to government involvement in private companies, but watching Anthropic’s handling of Fable makes me question whether they’re the company I want arriving at AGI/ASI first. The technology is one thing. The judgment and incentives behind it are another.
11
7
2
74
4,764
Derek Banks retweeted
I want to share a quick thought for people in cyber security. This will be my longest tweet ever. I’ve spoken to many lately who are having an existential crisis from the constant posts about “the end of cybersecurity jobs.” Yes, things are changing quickly. This is a significant moment for the tech industry. Change can be uncomfortable. But we’ve seen cycles like this before. • When GitHub and open source took off, people said software engineers would disappear because code was free. • When AWS and cloud computing emerged, people said infrastructure jobs would vanish. • When fuzzing and SAST tools improved, people said vulnerability research would disappear. • Virtualization would eliminate infrastructure jobs. • Mobile computing was going to end desktop dev. • Exploit mitigations would end exploitability. It didn't. Each time automation improved, the amount of software grew faster than the automation. It does feel "different" this time as it's explosive. Some roles will shrink: • repetitive pentesting • basic vulnerability scanning • tier-1 SOC monitoring But other areas are expanding rapidly: • AI system security • supply chain security • identity architecture • autonomous agent security • critical infrastructure protection Historically, every time we eliminate one class of bugs, new classes emerge. Right now people are vibe-coding entire systems, giving AI access to their machines, crossing trust boundaries, and deploying autonomous agents with excessive permissions. The legal and regulatory world is nowhere close to ready. There will absolutely be new failure modes. Humans are amazing and always adapt, finding new ways to do things. The worst thing you can do right now is fall into a doom loop. ...and I’ll be honest, I too have felt the "psychological paralysis" a few times thinking, “Is this time different?” It's especially impactful when it comes from someone I respect in the community. There are certainly unknowns, in an industry where we've become accustomed to predictability. But... the majority of those reactions are usually driven by social media, not reality. Platforms like X reward engagement, and sensational doom posts spread faster than measured thinking. If you see something like: “Holy #$%^! Opus 66.6 just found every bug in Chrome and replaced 50 startups!” …mute it and move on. Instead: Stay curious. Learn the new technology. Adapt your skillsets. Build things. We’ll get through this transition the same way we always have. If I'm wrong then Sam Altman better be right about UBI! :) I'm sure that if this tweet gets any engagement that I'll get some heat for it, but a good friend of mine reminds me often to focus on what you have control over. I'll revisit this tweet at DEF CON 40!
55
300
30
1,511
133,373
Derek Banks retweeted
I created a Claude and Grok integration into Signal running on my DGX Spark at home. I can have it create new projects, update code of existing products, test my code, do whatever Claude does as well as use the GenAI components of Grok to do research analysis for me. Only accepts messages from my number of course - it’s a linked account and accepts only from me.
14
6
109
10,466
Derek Banks retweeted
PAI is a super power. @DanielMiessler created features on top of Claude Code that increase its efficacy by 50%... and that's a lot based on how awesome Claude Code is. Incoming FREE workshop of PAI and other tools I'm using to modify how I work in cybersecurity. Straight out of our 3.0 update of our "Red Blue Purple AI" course.
Super hyped to announce PAI 2.3!!! A complete rewrite of the PAI system focused around: - USER, WORK, and SYSTEM data isolation - A Continuous Learning system based on hook-based Sentiment gathering - User-based Skill Personalization ...
9
12
1
122
10,591
Derek Banks retweeted
🔴 (Release) The Arcanum LLM Security Context Project Today we are releasing a new AI security resource for you all! This project is a comprehensive security reference distilled from 150+ sources to help LLMs generate safer code! Enjoy! 🫶 Links Below 👇
8
80
1
220
11,719
I think we all need to be full-stack now. AI generally, but Claude Code specifically, has made it so that if you can't go from problem to a shipped solution, you're just part of a toolchain that can/will be automated. Only high-agency, full-stack people survive. And they pull away from everyone else. That’s an insane change to happen over the course of like two months.
58
27
3
361
28,028