Veridise Re-poster
59 Posts
Veridise Re-poster
Hardening blockchain security with Formal Methods — Industry-leading security audits and tools for smart contracts, blockchain, and zero-knowledge.
0 Following
1 Followers
2 Liked
Posts
All
Videos
See translation
If you are curious about how $27M can exit a system through a simple reward calculation, @FormallyJon provides a clear breakdown of the situation. The true vulnerability was an underlying flaw in the accounting logic, which caused identical tokens to be registered two separate times. The system recognized the funds first as a deposit that the protocol was obligated to return, and then separately as an earned reward. Crucially, both of these balances were made available for withdrawal. While a reentrancy exploit functioned as the delivery mechanism for the attack, the flawed accounting was the actual bug.
See translation
Network upgrades on Mina can compel every zkApp to update its verification key, a demand that causes standard multisig setups to fail completely when trying to adapt. To resolve this challenge, Mina Multisig implemented an approach utilizing FROST. Before @nori_zk released this system on @MinaProtocol, Veridise conducted a thorough review of the solution, and this evaluation is currently documented on AuditHub. The software is now completely open source, making it an excellent resource for any Mina team working on wallet applications or self-custody tools.
See translation
The entire concept of being self-custodial fell apart as soon as individuals were required to deposit their assets into the spending contract for Rain. We witnessed a loss of $500K at Avici, along with $430K disappearing at Tria. In the end, any assertions about maintaining custody are completely meaningless if the routing contract you rely on happens to contain a vulnerability.
See translation
Asset managers and risk curators are now able to provide yield from just one deposit because the contracts from @Lombard_Finance accurately price Bitcoin deposits across various shards and converters. This setup allows for seamless vault migrations that successfully carry the correct share price forward. Prior to the platform scaling up with actual deposits, Veridise conducted a comprehensive review of the system.
See translation
It is completely possible to build a ZK circuit that passes all of your validation tests while actually computing an entirely unintended result. A new open source verifier called LLEQ has been created specifically to detect this exact issue, and it is officially live right now.
See translation
To identify bugs discoverable by artificial intelligence, a volunteer red team consisting of 20 to 25 developers has already examined the majority of the open-source software for Bitcoin. The reality today is that attackers no longer require years of expertise, as one cheap model is now all they need. Because of this rapid shift, an ad hoc review approach simply will not keep pace. Moving forward, provable guarantees absolutely will.
See translation
Panelists issued a warning this week regarding a significant shift in digital security. Historically, cybercriminals largely ignored a $20K wallet because the necessary effort far outweighed the potential financial return. The introduction of an AI agent completely eliminates that previous cost and benefit ratio. By leveraging this technology, a single attacker now possesses the capability to go after everyone simultaneously. Interestingly, the actual exposure was never the AI itself. The real issue is that our current wallet logic was specifically designed to defend against a human attacker.
See translation
In this week's edition of Auditor's Take, @FormallyJon highlights a fundamental principle behind every constant-product pool. These pools operate safely under a single critical condition, which is that external forces must never interact with their reserves. The moment that underlying assumption is compromised, the assurances provided by the pool collapse right alongside it.
See translation
As Ethereum developers gear up for the 2027 Hegotá upgrade, they are actively narrowing down a list of 66 proposals. While FOCIL has already secured its spot in the upcoming update, the most significant debate centers on specific EIPs. These proposals aim to equip privacy applications with a foundational layer that completely eliminates the need for trusted intermediaries. In the end, the chosen solutions will define the exact attack surface that every ZK auditor will subsequently be tasked with navigating.
See translation
In this week's edition of Auditor's Take, @FormallyJon breaks down the specific sequencing bug that recently drained Future Protocol. The core of the issue involved just three standard operations: a transfer, a sync call, and a burn. When these commands are executed in the proper sequence, everything operates exactly as intended and nothing unusual occurs. However, simply triggering those exact same steps in an incorrect order caused $4.6M to be emptied from the pool.
See translation
Executing a burn, a sync call, and a transfer in the proper sequence ensures that everything functions normally without incident. However, triggering these exact same three actions in an incorrect arrangement leads to severe consequences. This specific scenario is precisely how $4.6M was removed from the pool. For the latest edition of Auditor's Take, @FormallyJon provides a detailed breakdown of the exact sequencing bug that drained Future Protocol.
See translation
During a presentation at ETHCC, @FormallyJon shared a vital perspective on digital asset security. In 2024, exploits in smart contracts led to the loss of $348 million. A significant reason for these losses is that standard monitoring tools usually sound the alarm only after an intrusion has commenced, meaning the compromised funds have already vanished by the time anyone notices. We need to shift our strategy to identify these flaws proactively, rather than scrambling to fix vulnerabilities only after users have discovered them.
See translation
Recently, three dozen crypto firms submitted a request to AI labs asking to be equipped with the identical offensive capabilities that attackers already possess. We must keep in mind that achieving equality in search functions does not automatically equate to achieving equality in security assurance. Even though an artificial intelligence model can map out significantly more pathways than any human, it remains unable to verify the routes that it has never actually opened.
See translation
A math conjecture that stood for 87 years was recently refuted by Claude Fable 5. Interestingly, the discovered counterexample successfully clears the standard invertibility check. Because it routes three distinct inputs into one identical output, however, it remains impossible to invert.
This scenario clearly demonstrates that merely passing a verification check does not serve as an actual proof. It is precisely within this kind of logical gap that underconstrained ZK bugs exist.
This scenario clearly demonstrates that merely passing a verification check does not serve as an actual proof. It is precisely within this kind of logical gap that underconstrained ZK bugs exist.
See translation
Historically, security research within the ZK space has been highly compartmentalized. If you develop a specific utility for Circom, that exact same resource offers absolutely no value to Halo2 teams. Likewise, choosing to work with Noir means that Circom developers are completely left out. Because the industry has lacked a shared foundation to build upon, excellent research inevitably remains trapped inside individual environments. Fortunately, the introduction of LLZK is completely transforming this entire dynamic.
See translation
LLZK now has its own GitHub org, separate from Veridise.
Ethereum Foundation funded two rounds. Galois already contributed a backend independently.
Community projects built on the IR can live there. What would you want to see?
Ethereum Foundation funded two rounds. Galois already contributed a backend independently.
Community projects built on the IR can live there. What would you want to see?
See translation
Here is a bold perspective on the current landscape. The most significant unresolved challenge facing ZK tooling actually has nothing to do with performance. Rather, the fundamental issue is the lack of a common foundation. Because a universal base layer does not exist, each individual ecosystem is forced to construct its own security framework entirely from the ground up. I would love to hear your thoughts on this situation. Do you share this view, or do you believe that concerns regarding fragmentation might be exaggerated?
See translation
We are thrilled to share that LLZK V1.0 is officially available to the public. Designed as a shared intermediate representation, this framework specifically supports ZK security tooling. Developers will find that any programming language compiled down to LLZK instantly unlocks the capabilities of ZK Vanguard for static analysis and Picus for formal verification. The major advantage here is that you can leverage these resources seamlessly, completely avoiding the need to reconstruct either tool from scratch. As of this launch, our active frontends currently feature Halo2 and Circom. To dive deeper into this release, please visit the full announcement at https://veridise.com/blog/veridise-announcements/llzk-v1-0-a-new-phase-for-zk-shared-infrastructure/
See translation
While at EthCC, the single most popular question people asked was how formal verification measures up against AI auditing.
These two approaches actually seek to answer completely different questions. Artificial intelligence is designed to spot known patterns. Formal verification, however, provides an entirely different level of assurance by proving that specific properties remain true across every single possible input.
It is abundantly clear that this area of study has grown far beyond being just a niche topic. Solidifying its widespread importance, the Ethereum Foundation has officially committed $2M toward the advancement of formal methods.
These two approaches actually seek to answer completely different questions. Artificial intelligence is designed to spot known patterns. Formal verification, however, provides an entirely different level of assurance by proving that specific properties remain true across every single possible input.
It is abundantly clear that this area of study has grown far beyond being just a niche topic. Solidifying its widespread importance, the Ethereum Foundation has officially committed $2M toward the advancement of formal methods.
See translation
An overlooked issue is silently impacting the ecosystem of ZK security tooling today, namely severe fragmentation. Developers are currently forced to navigate an entirely unique tool stack for each distinct proving system and programming language they encounter. This reality means that foundational infrastructure must be reconstructed from the ground up over and over again. Ultimately, a development team's ability to successfully detect circuit bugs must never depend on the specific ZK language they decided to adopt.
Log in to explore more content