Chief Technical Officer @VeridiseInc! Building automated tools to secure the web3 ecosystems.
Somewhere in the US.
Joined September 2022
- Tweets145
- Following107
- Followers126
- Likes322
Kostas Ferles retweeted
To add on to this a bit, the point here is to discuss the potential impacts of quantum threats on current cryptographic systems. To make this as clear as possible, we'll spend less time explaining how quantum computers break cryptography and focus instead on the impacts
What if a quantum attack happens years before anyone realizes it?
@FormallyJon of @VeridiseInc will be at Devcon 8 to unpack “steal now, reveal never” attacks & why quantum risk may arrive long before quantum computers can openly break today’s cryptography.
Don't miss this one!
This incident is insane! From the lengths the attackers went to, to how it was detected, it shows just how drastically security and trust boundaries have shifted.
One sad reality is that if you're a @Ledger owner, you have to follow the news here to be aware of the compromise. I haven't seen any notifications through official channels.
The backdoored Ledger story keeps getting crazier. Attackers used the backdoor to grab $90+ million in crypto from users who bought hardware wallets precisely because hackers steal crypto keys from normal computers.
1) Someone built an implant into Ledgers to steal keys, with an “LTE module, Antenna, e-SIM, a MCU connected to the Ledger’s SPI bus.” (c/o @MagicalTux)
2) It looks like the attackers acquired CryptoBilis, a Ledger authorized retailer, and tampered with their inventory.
3) @MagicalTux found the implant by accident. He was tearing down Ledgers for unrelated reasons and bought them from the cheapest place possible, which was, unsurprisingly, the malicious CryptoBilis.
4) Once the implant was detected, the attackers quickly used the keys they’d stolen to drain the wallets.
5) There appear to be multiple generations of implants.
We know that the asymmetry between attackers and defenders has broaden since the beginning of this year. That impacts any tech firm regardless their size, not only big tech. This is a good time for everyone to reassess their security priorities. Happy to chat if you are looking to harden your security in a any way!
Kostas Ferles retweeted
Happy to have shared at the Solana Summit today our new Solana wallet support for Midnight!
@solana users will be able to do private swaps and interact with private DeFi directly from their Solana wallets (no separate install required!)
Kostas Ferles retweeted
1/ The next stage of Midnight is here.
Permissionless smart contract deployment is now live on Midnight mainnet.
That means builders can deploy contracts directly without additional security reviews, while contracts can also dynamically deploy other contracts on behalf of users.
Interesting, one more dimension where we gonna have multiple models to choose :)
If true, the era of AI subsidies is coming to an end. Let's start discussing how we can use tokens/inference more efficiently :) Get the most value out for what you pay!
OPENAI 🔥: The upcoming ChatGPT Pro Max plan will cost $500.
So far, this will be one of the most expensive AI subscriptions available on the market.
I hope "it will be worth the wait" 👀
Glad to see that formal method techniques are becoming mainstream.
However, I wished things were as simple as saying to an LLM "verify this." Verification is a more thorough and rigorous process, where you need to clearly define what you need to prove. Leaving that up to the LLM tends to result in poor coverage and the software might still contain critical bugs. And when this happens these days, it's usually formal methods that takes the blame.
Even in this post, the main output of the chat was how many bugs were found and fixed via lean. But the most interesting part when you are using lean is *what properties* you proved.
I used Opus 5.5 to formally verify the Claude Agent SDK using Lean. A couple short prompts = 16 PRs fixing various bugs and race conditions. Video attached.
TLA+ also works well. I sometimes combine Lean and TLA+ to look for issues around data flow, concurrency, and state mgmt.
I don't know either language well, but Claude is excellent at both. This approach is super useful for formally modeling your code and finding bugs that a human probably wouldn't have spotted.
Is formal verification the future of coding (or at least, bug finding)?
Sad but true. Apart from a marketing tool, "rogue" agents have also become a proof of LLM cyber capabilities. I can't wait for the hype to be over and the actual work to start. As white hats, we must be pointing the agents on the other direction and help secure projects!
Kostas Ferles retweeted
A function that forwards caller-supplied calldata into a low-level call can perform any operation the target's interface exposes.
Nobody designed that set. No auditor reviewed it.
New Auditor's Take from @FormallyJon covers the class.
veridise.com/blog/audit-insi…
Completely agree! Although, creating a reliable and consistent defensive agentic solution has become a weird art form. You need to jump through hoops like bypassing cyber restrictions (imposed by the frontier itself), which will make this an uphill, but interesting, battle.
Kostas Ferles retweeted
Formal verification tells you a property fails. It doesn't tell you why.
@ShankaraPailoo2 on turning a counterexample into a root cause.
Kostas Ferles retweeted
For cybersecurity, "pacing the frontier" means "policing the frontier".
Frontier cybersecurity solutions will be driven by defensive agentic AI.
I read @DarioAmodei's essay calling on the labs to pace the frontier. @sama agreed. The frontier will move at whatever speed it moves. The rest of the world will not slow down. Our job in the cybersecurity community is to make sure it moves securely and safely.
Here’s what I see from the front lines:
1. The unit of threat is no longer the hacker. It's an autonomous campaign. I call it the Agent-state. We see coordinated AI agents executing attack campaigns at machine speed.
2. Sophistication is dead as an attribution signal. AI gives every criminal and lone actor elite execution. Identity, infrastructure, and intent tell you who's behind an attack. Skill doesn't.
3. Runtime is the control point. Endpoints, cloud workloads, and SaaS are the battleground. Governance documents don't stop an agent in motion. Enforcement at machine speed does.
4. Every AI agent is a privileged identity. Least privilege, short-lived credentials, traceable actions, and a kill switch. Permissions never expand because an agent decides it needs more.
5. Defense has to be autonomous but also bounded. Machine-speed response, tiered by consequence, with humans owning the high-impact calls.
6. Every failed attack should make every defender smarter. Feed what we block back into detection, across customers and models, with privacy intact. This is what CrowdStrike and NVIDIA introduced with SafeMind: an agentic, always improving model and harness protection system built for defenders.
7. The AI industrial base is critical infrastructure: weights, training clusters, APIs. Call it what it is and protect it like it is.
Pacing what comes next doesn't secure what's already here. The credible path is to deploy with proof: board-level accountability for AI security, independent external red teaming, incident disclosure, secure defaults, and controls that work in production - not on paper. Anthropic and OpenAI just committed to embedding independent evaluators with employee-level access. CrowdStrike will bring what we see from the front lines to that table.
The ability for AI to act must be matched by the ability for defensive AI to stop the breach.
Terrifying in two ways:
1. how people are weaponizing AI
2. how LLM providers can trace what you do
I get the latter for malicious actors, but to what extend this happens for "regular" users? Regardless, I think solutions like @usewisp_io should be the ultimate way to go!
We're publishing our most detailed threat intelligence report to date.
It covers how people tried to misuse Claude—for cyberattacks, influence operations, surveillance, biology, and building weapons—and how we found and stopped them.
We disrupted every operation in the report, and used the lessons from them to strengthen our safeguards. Where appropriate, we also shared what we found with authorities and other AI companies.
These cases are not typical: we’re highlighting some of the most sophisticated misuse we’ve seen. But they’re especially important to discuss, because they show us where AI misuse is headed, where our safeguards work, and where they need to improve.
We’re publishing this report so others can spot the same activity on their own platforms, and so we can give the public a clearer view of how emerging threats develop.
Read the report: anthropic.com/threat-intelli…
Mostly true for exploits today. AI can handle writing a PoC/exploit after you found something. For vulnerability discovery, we still have some skin in the game (even if it's just improving prompts and/or threat models).
Kostas Ferles retweeted
Love seeing @ZKPassport and @TACEO_IO ship this. Noir proves the passport predicate on the phone. A threshold OPRF stops one issuer or operator from reproducing the app-specific nullifier. Voting, access control and rate limits get uniqueness without an identity graph.
New: unlinkable on-chain identity.
You can now prove what's necessary on-chain, with no one, not even the authority that issued your passport / national ID, able to trace it back to you.
It now runs on @TACEO_IO 's oblivious network.
Sybil-resistant. On-chain. Unlinkable.