ZoA 🦀 (@zoasec) header
ZoA 🦀 (@zoasec) profile picture

Breaking blockchains and smart contracts to make them unbreakable Resident and Judge @cantinaxyz TG: https://teapawt.net/t.co/Gusj1GKKTx

Blockchain
Joined December 2024
Photo by ZoA 🦀 (@zoasec)Photo by ZoA 🦀 (@zoasec)Photo by ZoA 🦀 (@zoasec)Photo by ZoA 🦀 (@zoasec)Photo by ZoA 🦀 (@zoasec)Photo by ZoA 🦀 (@zoasec)Photo by ZoA 🦀 (@zoasec)Photo by ZoA 🦀 (@zoasec)Photo by ZoA 🦀 (@zoasec)Photo by ZoA 🦀 (@zoasec)
Another story has been written: the @GammaStrategies competition standings are in. 🪐 Your top-ranked researchers: 🥇 @ZoAsec: $12,500.00 OP 🥈 @0xOllam: $7,500.00 OP Thanks to everyone who participated. Full leaderboard below.
3
4
1
35
2,711
I aim to have 50+ public Solana security reports, to be a reference for anyone who wants to enhance the security of their Solana program, learn Auditing to become a Security Researcher, or even become a Blackhat. Bookmark this to stay updated: github.com/Frankcastleaudito…
9
11
121
5,049
Which rug pulls, exploits, and security breaches happened this week? Read this 2-minute weekly summary to stay in the loop 🧵👇 (1/6)
6
1
19
6,901
1/ Security isn’t one-size-fits-all. At Fuzzland, we built Blaz+ to cover every surface attackers can hit — from smart contracts to social media. Built for founders, by hackers. Here’s what full-spectrum coverage looks like 🧵👇
3
6
1
20
2,865
🚨MistTrack Security Alert🚨 @zothdotio has been exploited, losing ~ $8.32 million. 1⃣Funds are currently at address 0x7b0cd0d83565adbb57585d0265b7d15d6d9f60cf. 2⃣The initial source of funds from the hacker address 0x3b33c5cd948be5863b72cb3d6e9c0b36e67d01e5 is #ChangeNOW.
🚨SlowMist Security Alert🚨 We have detected that @zothdotio has been exploited, likely due to a leakage of Admin privileges, resulting in the logic contract being tampered with and replaced by a malicious contract. Btw, thanks to @0xtroll for the shout-out. As always, stay vigilant! etherscan.io/address/0x82f3a…
1
5
1
10
7,166
⚠️A critical vulnerability (GHSA-vjh7-7g9h-fjfh) has been discovered in the widely-used elliptic encryption library. 😈Attackers can exploit this flaw by crafting specific inputs to extract private keys with just a single signature, potentially compromising digital assets or identity credentials. ✍️In our latest article, we break down the vulnerability—its root cause, impact, and how to mitigate the risks. ❤️Special thanks to @Rabby_io for providing the vulnerability intelligence. 🔗Read the full analysis here: slowmist.medium.com/private-…
9
35
14
132
24,408
ZoA 🦀 retweeted
For $4.7 billion dollars a year (the total SOL emissions) we can definitely make it a ZK rollup for far cheaper. 1000 TPS * 60 * 60 * 24 * 365 * 1/10 of a cent per transaction proving cost = $31M a year
18
12
3
198
31,212
ZoA 🦀 retweeted
How to run a Proof-on-Concept on @Stacks 🛠️ Find bugs and earn from their $250k reward pool #StacksxImmunefi
2
4
14
1,507
Previously, I said I was working on the Manhattan Project of blockchain security. Today, we unveil it. It’s called Magnus: the first platform to unify the onchain security tech stack into a single command center. We believe its industry-wide adoption will cut hack impact by another order of magnitude. With the devastating Bybit hack, having a SecOps command center is no longer optional. Here’s how Magnus makes security robust and efficient for your project in a comprehensive new way.👇
6
21
2
97
12,535
ZoA 🦀 retweeted
This is Pectra. Ethereum next major upgrade. In 2 month, it will cause new wave of hacks. Here’s what you need to know (and how to prepare) 🧵 1/6
15
42
4
214
23,106
ZoA 🦀 retweeted
#DailyWeb3Security News – 18.01 - 19.01 Been deep in bug bounty hunts or audits? Here’s your quick catch-up: The biggest security updates, hacks, tools, and trends from the past 24 hours. If this helps, like/retweet to keep the knowledge flowing!
2
1
3
337
Great visual representation of how Solana works 👇 report.helius.dev/
1
9
36
1,665
Web3 Hack Postmortem 2024 We are excited to announce that we are releasing our annual report again, with 330 pages this year! This report includes: → Statistics of 204 exploits: scale, root cause, ... → Correlation between security audits and hacks. → Post-incident responses from the team → Destination of stolen funds → Standardized Process on Security Incidents → Individual case analysis This year, we open the report in public from its first release. Also, we would like to express our special thanks to @_SEAL_Org, @isaacpatka, @pcaversaccio who have been actively securing web3 ecosystem and helped writing the report! For accessibility, we provide the shortened version (without case studies) through the medium article. PDF version of the full report is attached at the article as a google drive link. 👇
4
23
6
80
14,743
just published our 2025 mission, vision, & strategy this is the year we build together — give it a read
168
105
59
960
125,875
The Ultimate Security x AI Agent Guide > Do we need AI agents in crypto security? Are AI agents just another VC meme? It all starts with the existing fragmentation in security: There are over 200 active audit firms[1] and >2,000 active independent auditors in crypto security. > Why is this the case? It exists because there's no “right” way to secure a codebase. Any method that results in finding vulnerabilities is legitimate. And no single method can provably find every vulnerability. These are facts. To complicate things further, many vulnerabilities are kept private. So even if you built a machine that could find every publicly disclosed vulnerability, you’d only have protection against a fraction of the possible vulnerabilities. > Where do AI Security Agents fit into this? Recently, there’s been a trend of “cross-checking” answers across different LLMs[3]. ChatGPT gives you one answer, but Grok gives you another, Claude, Perplexity, Gemini give still different answers. Each model creates an answer that, like a snowflake, is unique. This is because each model is architected differently, trained on different data, and fine-tuned differently. In the same way, a security AI is trained on unique data in a unique way. So it will give a unique output compared to another security AI. Some security AIs will have access to non-public information. These AIs may then be capable of finding vulnerabilities that others can’t. Therefore, a "swarm" of AI agents has a high chance of finding more vulnerabilities than one of those agents acting alone. > But what if an AI achieves AGI? Is the swarm irrelevant? AGI measures an AI’s ability relative to humans. As we’ve learned countless times in the security space, one human (or even a brilliant group) cannot always find every vulnerability in a codebase. This is why auditor teams, audit contests and bug bounties have become so popular: you get more than one pair of eyes looking at your codebase. There's a very big surface area where AIs reach AGI (become better than humans) but still miss bugs. AGI does not mean "perfect". And if any single AI has a chance of missing bugs, then a group of AI agents would be preferable due to the greater likelihood of finding all bugs. Even in a scenario where a perfect, all-knowing security AI exists, it will still be difficult to prove that it's perfect and all-knowing. So having multiple AIs would still be the safest path. Right now, having multiple security researchers audit a codebase is better than one. The point is this: Having multiple AIs securing a codebase will almost always be better than one. Groups of AI agents will be the meta in crypto security. > But what if AI agents never get as good as human security researchers? We've covered the "omnipotent" scenario. Now let's discuss the "useless" AI scenario. AI can already find critical vulnerabilities in code. It's still in early stages and the codebases need to be fairly small. But there are dozens of examples of AIs finding bugs in code[4]. Reentrancy bugs, safe transfer bugs, you name it. Here's why that matters: An AI that can find even 10% of the vulnerabilities that a top security researcher can find is still very useful. There are two huge advantages that an AI has over a human security researcher: 1) AI can audit a codebase 100x faster (likely 1,000x) 2) AI can audit a codebase 100x cheaper (likely 1,000x) This means you can run the AI during your development cycle. Merged a PR? Great, run the AI. Integrated a new library? Great, run the AI. Deleted one line of code? Great, run the AI. Even if it can only find 10% of the vulnerabilities, its cost and speed make it a very useful tool. > What will these AI security agents help with exactly? Non-Smart Contract Use Cases 🐠 Phishing detection An in-browser security agent can detect patterns in messages to alert a user of a possible phishing attempt. 🦠 Anti-virus An AI agent can scan file downloads to detect patterns commonly associated with malware. 📐Transaction tracing An AI agent can quickly trace through thousands of transactions to reveal relationships between wallets and funds transfers. 🟥 Phishing Red Teams AI agents can personalize fake phishing messages to harden a company's employees against phishing attempts. Smart Contract Use Cases 🧑‍💻 Smart Contract Auditing We know AI agents can find bugs in smart contracts. The main questions are how good they’ll get and how long it takes them to get there. 👨‍⚖️ Vulnerability validation (Judging) AIs can hallucinate millions of vulnerabilities and write convincing code to "prove" them. A separate "judging agent" can validate whether a submitted vulnerability is real or not. 🧪 Test writing Tests play a large role in validating code paths and preventing exploits. An agent that can write useful tests will save developers hours of mind-numbing effort and keep protocols safer. ⛽️ Gas Optimization Some AIs are already quite good at gas optimizations because many optimizations fall under a few categories that an AI can easily learn and apply to new codebases. 🤵‍♂️Formal Verification / Invariant Writing Imagine this: You give an agent a 1-page description of your lending protocol, and it writes 75 invariants about your protocol. Some of the invariants may be wrong, but validating existing invariants is much easier than coming up with them. 🍑 Fuzzing Fuzzing is where you throw random inputs into a code path until something breaks. An AI agent could enable “smart fuzzing” where the inputs are less random and designed to break a function/flow faster. An AI agent can also decide which code paths to fuzz in the first place. 🖥️ Monitoring AIs can monitor the blockchain for unusual activity in the hopes of finding (and blocking) something malicious. Monitoring and threat prevention is a "cat-and-mouse" game, and AI agents can better detect transactions trying to use loopholes. 🛳️ Deployment Audits Remember the $10M bug bounty paid by Wormhole due to an uninitialized proxy contract? AI agents can help catch those errors and ensure a protocol deployment goes smoothly. 📚 Documentation Writing Imagine NatSpec but smarter. You write the code and some comments, the AI agent turns that into full-fledged documentation. It's much easier to edit existing documentation than write it from scratch yourself. 🪙 ERC-20 token audits A subset of smart contract auditing that will likely become 99% AI agent-driven soon. Now you pay $10k and it takes 1-2 days. Soon you'll pay $100 and it takes 1-2 hours. > How should security researchers approach the advent of security AI? Check out my article here on the 3 steps to take: teapawt.net/jack__sanford/status/1… > What are the most promising AI security agents? Here are my current top 5: teapawt.net/jack__sanford/status/1… Citations in reply.
Top 5 AI Security Agents (January 14th, 2025) Based on some early research, these security agents seem to have the most mindshare and at least a small track record of finding vulnerabilities. 1) Audit Agent by @NethermindSec I've only heard good things about this agent and I know a lot of people who have tried it. Interface is very easy to use. Its business model (showing you the top vulnerability but you pay for the rest) also exudes confidence. 2) Certaik @CertaiK_Agent Listed on @virtuals_io with a $5M market cap. I've met the founders and they seem technically strong. It has discovered some findings in ERC-20 token audits which is quite easy to be fair. And Certaik is currently participating in the Aave V3.3 audit contest on @sherlockdefi. 3) H4CK Terminal Listed on @virtuals_io with an $11M market cap. It's not an audit agent per say, it's a white hat agent. It has found and been paid for some frontend vulnerabilities (including in @virtuals_io itself) which is fairly impressive. Team seems to be full-time on it. 4) HOUND by @muellerberndt This one makes the list because @muellerberndt is a legend in the space, having been an early auditor at Consensys Diligence. The approach seems to be more task-oriented like a true agent. And the output looks nice. 5) SolidityScan by @CredShields This one seems to be a more full-featured product with a multi-screen UI. I personally know less about the team and haven't tried it yet. If you give it a go, please let me know! solidityscan.com/ I'm sure this list is incomplete, so please share any other security agents you come across and I will feature them next week ✌️
9
6
3
36
5,486
ZoA 🦀 retweeted
👾😎Congrats to @KeiZo_Zo who has won himself an exclusive SR bathrobe! Stay tuned for our next swag drop coming on 20/01!!!
👾 🎁 Are y'all ready for the first Swag drop of 2025? Rules are simple: - Submit a NEW report AFTER this tweet - Say "Swaggity Swooty" in your report - RT this tweet! - Get confirmed 🏆 Crit = Bathrobe 💎 High - Holographic T-shirt 📘 Medium - Notebook 📢 Low - ???
1
4
16
2,423
ZoA 🦀 retweeted
2024 was a record-breaking year for onchain security, and we’re proud to have played a part 🛡️ We reached $112M in total payouts, our community found 4,100+ bugs — and kept 500+ projects secure. 💪 Here’s to raising the bar even higher for onchain security in 2025! 🚀 Want to work with Immunefi? Get started here: immunefi.com/projects/get-st… #Immunefi2024
6
22
420
744,880
Morpho is the fastest growing lending protocol on @base. A quick recap of its growth since deployment 6 months ago 👇
17
34
7
205
25,203
Big moves this week at @CertaiK_Agent—constant shipping ahead! 🚀 I’m especially excited to see how CertaiK can empower auditors in the @aave contest on @sherlockdefi It's only the beginning.
1
6
31
1,322