John Connor (@ConnorSecurity) profile picture

Cybersecurity from the future

Joined December 2019
Photo by John Connor (@ConnorSecurity)Photo by John Connor (@ConnorSecurity)Photo by John Connor (@ConnorSecurity)Photo by John Connor (@ConnorSecurity)Photo by John Connor (@ConnorSecurity)Photo by John Connor (@ConnorSecurity)Photo by John Connor (@ConnorSecurity)Photo by John Connor (@ConnorSecurity)Photo by John Connor (@ConnorSecurity)Photo by John Connor (@ConnorSecurity)
John Connor retweeted
Announcing a major milestone achieved by my team @GoogleDeepMind: we have proven key security properties of the machine code underlying the Signal messaging protocol. This is the first time the actual production code of a widely deployed protocol has been formally verified – by any method, with or without AI – and at this depth. More than a research breakthrough, it’s a prime example of how frontier AI can be a force multiplier to counter the speed and scale of modern cyber threats. Formal verification offers a higher standard of security; however, it’s extremely complex and time consuming, which is why in many cases it’s considered to be impractical. Similar efforts could take teams decades of combined effort. We did it in a month’s time. Using AlphaProof Code, Google DeepMind's new agentic framework for formal verification that runs on LLMs including Gemini Carbon, we verified the Signal protocol down to the actual code running on devices.  This milestone lays the groundwork for a future in which vital systems, from medical device firmware to power grid software, ship with mathematical proof that their protections work. Reaching that future will take continued research from our community. In that spirit, we're open-sourcing our formal specifications and verification artifacts so others can build on this work. On a personal note, this moment is of deep significance to me. I started my own research journey more than two decades ago as an intern working on checking the consistency of complex software systems. Seeing this grand challenge made practical with modern AI is extraordinary. I'm extremely proud of my team members: Swarat Chaudhuri, Andrew Ferraiuolo, Anton Kovsharov, Tej Chajed, Sam Schlesinger,  Karn Seth, Jonathan Katz, and Charles Sutton for taking on this grand challenge and delivering, and grateful to the Signal team for their partnership. deepmind.google/blog/formal-…
234
701
239
5,841
607,550
John Connor retweeted
everyone should read and realize this for once atleast
37
1,130
77
11,847
214,559
John Connor retweeted
People focus on the names, on the frameworks … forget the names. They’re all smoke and mirrors. Anyone could task a coding agent with studying one of these frameworks and building a custom version in another language, with a different flavor, skill set or type of orchestration. We’re in a new era now. Implementations are becoming disposable. Thousands of lines of code can be spewed out with very little effort. The principles and practices matter more than the framework names, because those ideas can materialize in code after a single prompt.
33
34
2
204
30,919
John Connor retweeted
Now that we can point machines at everything we’ve built over the last 40 years, I’m starting to think a lot of it was “security by obscurity” all along. There just weren’t enough skilled people looking at how broken it actually was.
36
32
6
367
21,350
"zero-day malware" is not a thing. Whether or not an EDR will detect any given piece of malware depends entirely on how well the security team configured it. Threat actors are not going to deliberately drop malware that is detected by your EDR, because that would be extremely pointless (it would just get automatically removed). The choice of terminology is a very deliberate attempt to make you associate it with "zero-day exploits" (novel unknown vulnerabilities in software), then come to the conclusion that they were hit with some completely novel unstoppable attack. In reality, zero-day exploits are also detectable with good security configurations, but this wasn't a zero-day exploit, it was just normal malware, which is detectable with a well-configured EDR. Even if the malware was brand-new, never before seen, coded from scratch, a good EDR configuration can and will detect it. The entire purpose of EDRs is to detect malicious behavior, not how the malware looks or whether or not it's "new". You don't lose $6 billion dollars as a result of a compromised developer endpoint unless your security posture is extremely poor. They are trying to cover their ass with disingenuous wording choices.
Replying to @MalwareTechBlog
umm the fake interview assessment installed 0-day malware that went undetected by EDR.
14
25
221
10,974
John Connor retweeted
The cost of not being curious is the highest right now. We no longer need devs who just tick off tickets from their boards. Agents have gotten pretty good at it. What we need are curious people. Who are willing to go into a rabbit hole on weekends, learn something new, experiment, break things and come back knowing something. The good thing about LLMs are they have made this entire process a lot simpler and faster. There's literally no more friction in learning something new. Funny part is curiosity is also something that makes u good with agents. Your output only depends on how well u explain the thing and you cant explain something clearly you were never curious about.
4
27
1
141
5,746
John Connor retweeted
High priority: AI Firewall AI Endpoint Protection AI SIEM AI Identity Low priority: Firewall Endpoint Protection SIEM Identity
In a single year, AI has rewritten the security priority list for CISOs. Attackers are moving faster. Bain estimates AI compresses a typical cyberattack from about four weeks to about 18 hours, and models keep getting better at finding and exploiting weaknesses. There's also more to protect, with employees shipping apps they couldn't have built a year ago and agents getting access to source code, documents, and production systems. In a survey of 100 CISOs at companies with 1,000+ employees, all seven of the most-selected 2027 priorities involve AI. The biggest jumps since 2026: - Internal AI agent security (+28 pts) - AI SOC (+26 pts) - AI offensive security (+24 pts) - AI code security (+23 pts) - AI vulnerability remediation (+22 pts) Last year's staples: - Exposure & vulnerability management (-21 pts) - SIEM / security data (-18 pts) - Privileged access management (-18 pts) From a16z's @OwenColby and @steph_zhang on The New Security Agenda: a16z.news/p/the-new-security…
14
22
5
264
17,897
John Connor retweeted
AI has really reduced the friction for novices to create a lot more work for experts. This php2shell event is just a widescale example of it.
8
10
84
4,165
John Connor retweeted
Well that's it for FLARE, here are the top human finishers. Appreciate everyone who participated in our little experiment, thanks! I think we will look back on this as a watershed moment for RE, impossible to ignore the future. The future is now 🚀
5
3
68
6,894
Malware is cool because either 1) If you're writing malware: you're programming with the intent to be as convoluted as possible to evade detection. It is a fun programming challenge. 2) If you're reverse engineering malware: you're taking apart something someone invested time into to be malicious or convoluted as possible Bonus points are you learn a lot about the internals of stuff (Operating Systems, file formats, network architecture, etc) and, believe it or not, people are just distributing MALWARE for FREE. Malware is all over the internet and you can just TAKE IT and no one cares
52
135
6
2,310
51,690
John Connor retweeted
Why would I pay $15/month for your software when I can plan, build, review, improve, host, launch, scale, secure, patch, monitor and debug it myself?
460
383
111
5,832
217,633
John Connor retweeted
Reverse engineering has become a commodity. Agents can do it at warp speed, while the human role is gradually shifting toward deciding what to examine, validating the results, and deciding what to do with them.
12
16
3
208
11,034
John Connor retweeted
We don’t get to worry about “alignment” for human threat actors. We never have. We have never relied on the threat actor to do the right thing. We have used disciplined security control systems to simply prevent their access.
8
50
10
257
11,141
Everyone chases custom APT malware; what actually ransomed your network was a $10 stealer log, an unpatched VPN appliance, or an open RDP port. Custom malware is custom for a reason - often tailored per target, and that target probably isn't you. Most orgs would get more out of hunting commodity than chasing APT tooling.
19
26
5
216
10,496
Worried AI will take your job in cybersecurity? After this week, I'm not. I've spent 2+ days reversing a single malware sample, and I'm still going. AI reverser skill did a lot of the heavy lifting for me, but the judgment calls are still mine, and those can't be rushed. So now I run on two brain cells: one that verifies everything the AI tells me, and one that keeps training it to do a better job next time. AI is becoming the engine of malware analysis and threat research, and it gets better every time a human checks its work and corrects what it missed. The future isn't AI or analysts. It's AI WITH analysts. Trust.
23
20
1
277
12,941
John Connor retweeted
1
44
4
193
7,971
John Connor retweeted
Out of the hundreds of popped organizations, the adversary was only able to get domain admin on 12 victims. Where domain admin was achieved, it was achieved in break neck speed. The takeaway there is hardening DOES matter.
5
8
3
79
17,720
John Connor retweeted
Hundreds of AI agents powered a campaign against PaperCut NG/MF. What did the operation reveal about agentic attacks and their limits? 🔗 Read GreyNoise’s latest blog: greynoise.io/blog/ai-orchest… #GreyNoise #ThreatIntel #CyberSecurity
13
5
48
19,172
John Connor retweeted
The recording of my second Binary Cartography webinar is public: Agentic Malware Analysis: From Task Automation to Deep Analysis Topics: string decryption, API hashing, unpacking & pipeline building Recording: youtu.be/azej1P17w9E Slides & samples: github.com/mrphrazer/binary-…
2
90
298
16,763
For most of 2025, I was skeptical that AI was already playing a major operational role in real intrusions. Most public examples seemed limited to phishing and supporting tasks. This report by my friend Eyal Eyal lines up with what I have been hearing elsewhere, too - in recent publications and in private conversations with people seeing this stuff up close. I think that phase is over. AI is moving into the operational core of attacks. With stronger models, open models, and jailbroken variants circulating, the economics have changed. Tailored tooling, exploit adaptation, and large-scale analysis get cheaper and faster. I expect AI to play a major role in future campaigns, and that means more variation, more fresh tooling, and less reliance by attackers on recycled code. All the more reason to focus on controls and detections that do not depend only on known samples. Worth reading.
8
56
3
285
50,504