Perimeter (@perimeter_sec) header
Perimeter (@perimeter_sec) profile picture

Industry leading security services specialized in fuzzing, securing DeFi's largest protocols

Joined January 2024
Photo by Perimeter (@perimeter_sec)Photo by Perimeter (@perimeter_sec)Photo by Perimeter (@perimeter_sec)Photo by Perimeter (@perimeter_sec)Photo by Perimeter (@perimeter_sec)Photo by Perimeter (@perimeter_sec)Photo by Perimeter (@perimeter_sec)Photo by Perimeter (@perimeter_sec)Photo by Perimeter (@perimeter_sec)Photo by Perimeter (@perimeter_sec)
Perimeter retweeted
Awesome resource for EVM Fuzzing, every Solidity developer should study this, do not skip your fuzz testing Tens of tools and solutions, Fizz from pashov/skills also mentioned🙏Thank you for compiling this list and the contribution @rappie_eth 🫡 github.com/perimetersec/evm-…
6
29
216
9,830
Shout out to @rappie_eth with @perimeter_sec for finding the above example. He's been fuzzing behavior differences in various solidity compilers. 3/3
1
2
21
1,376
Perimeter retweeted
Found this through differential fuzzing and first thought it was critical. Turns out --via-ir can change a function's behavior, as documented among the breaking changes (link below). Sharing this because it surprised me, and will probably surprise others as well.
Most people don't know that the --via-ir flag makes changes in the Solidity language. If using it to compile with, your code can do different things from compiling without it. If using it, here's some things you should do: 1/3
1
4
25
1,789
I received a lot of questions after the @perimeter_sec announcement I am still providing my services as an individual to select clients, but I am also happy to work with teams that share the same values Perimeter is a curated fuzzing group with a work-based model, similar to @spearbit, @zenith256 or @AdevarLabs I joined because I believe invariant-driven development is a key step toward improving the industry's security maturity, and because they share the same focus on quality and mindset If you are passionate about fuzzing, you should join their Discord (you can ping @rappie_eth for an invite)
1
4
37
2,151
Looking forward to working with the Perimeter folks and continuing to push the state of fuzzing!
We’re excited to welcome Josselin to Perimeter as Lead Fuzzing Specialist! Ex-Engineering Director of the blockchain team at Trail of Bits. Author of Slither. 🧵 highlighting some of his achievements
11
4
82
4,461
Which fuzzer should you use? People often ask me this question for their codebase, I've created a handy flowchart to answer this question! Here's the reasoning behind my recommendations: Stateful fuzzing: Use Echidna and Medusa by default. They’re the most reliable and have different heuristics, meaning one might catch an issue the other misses. Stateless fuzzing: Foundry is the best choice due to its strong support and seamless integration into a Foundry test suite. Slither is OP: Echidna heavily relies on Slither for value generation. Without it, it’s like using a machine gun with an empty magazine. On the other hand, Medusa is designed to work independently of Slither but now supports it for value generation. ffi breaking coverage: When using ffi in other languages, fuzzers struggle with coverage guided fuzzing. In these cases, prioritizing speed with Medusa and Foundry is more effective. On-chain fuzzing: Medusa isn’t fully recommended yet for this, as it’s still experimental, but that could change soon! These recommendations are current, but can change in the near future with new updates. For example, Medusa just reached the v1.0.0 milestone. Let me know if you agree or disagree! PS: Did not cover other fuzzers like ItyFuzz and Wake since I haven't personally tried them out enough to give a conclusive opinion
10
8
3
74
9,441
How Can Protocols Potentially Save Millions in Their Security Budget? The key is incorporating invariant testing with fuzzing early in the protocol development lifecycle. Here’s how it significantly cuts costs: 1. Reduces Costly Follow-up Audits Without sufficient testing, initial security audits often uncover severe issues that require major modifications and architectural changes. This leads to expensive follow-up reviews. By implementing invariant testing with fuzzing early, many of these issues can be identified beforehand, saving 5–6 figures in additional audit costs. 2. Increases Coverage and Reduces Bug Bounty Payouts In addition, fuzzing detects different types of vulnerabilities compared to manual audits. Combining both approaches significantly improves security coverage, reducing the likelihood of high-severity exploits and potentially saving 5–7 figures in bug bounty payouts, or worse, being exploited. 3. Improves Manual Audit Efficiency Since fuzzing is more suitable to detect issues such as rounding errors, manual reviewers can focus on vulnerabilities that are more suitable to detect through manual analysis. This improves audit efficiency and delivers 4–6 figures in added value. 4. Continuous Security Throughout Development Invariant suites persist across all code versions and can be updated as the protocol evolves. This allows continuous re-testing to detect vulnerabilities introduced by code changes. 5. Verifies Fixes and Prevents Regression Issues identified in security reviews can be added to the invariant suite to verify that fixes are effective and that no new vulnerabilities are introduced. This process can potentially save 4–6 figures in future remediation costs. 6. Strengthens Code Robustness Incorporating a thoroughly made fuzzing harness encourages the protocol developers to think more critically about protocol invariants, leading to a more resilient codebase. Many of these invariants can be reused in future protocol version, improving security and efficiency long-term. The combined savings and added value to the protocol can easily reach 5 figures for smaller protocols and scale to 7 figures or more for larger ones! Of course, the benefits only applies for fuzzing harnesses with proper invariants and extensive state exploration. Based on what I’ve seen and from personal experience, every protocol with a well made fuzzing harness has gained significant value from it and has never regretted the investment. For a practical example of an invariant fuzzing suite, check out: github.com/perimetersec/zkev… For more examples check out: github.com/perimetersec/publ…
1
2
1
20
1,533
Happy new year's everyone! 2024 was an extremely successful year for myself and @perimeter_sec! We have some very exciting things to announce + share early in 2025 Keep your đź‘€ peeled, here's a quick hint:
1
7
683
Appreciate the response, and not shying away! I believe it's important to have these discussions transparently and publicly to advance the space as a whole. Below I respond to your arguments, and provide a simple solution to your challenge đź§µ1/n
1/9 Thank you for continuing the discussion. We have answers, and as a bonus also a fun challenge for everyone đź§µ
5
8
2
46
9,570
The Echinda fuzzer can optimize. We had code that we wanted to see how bad things could get with an important safety feature disabled. Rappie used optimizing to determine the max sizes for possible accumulated rounding errors. Was very useful. Prod still has the safety. :D
Excited to share the latest fuzzing project with @perimeter_sec. We focused on @OriginDeFi's recent refactor to simplify their Vault contract. 22 invariants tested for 130M+ runs 1 high severity issue found 12 tolerances established and tested Many thanks to @danielvf for his support during this engagement. Links below👇
1
3
25
3,817
Perimeter retweeted
Excited to share the latest fuzzing project with @perimeter_sec. We focused on @OriginDeFi's recent refactor to simplify their Vault contract. 22 invariants tested for 130M+ runs 1 high severity issue found 12 tolerances established and tested Many thanks to @danielvf for his support during this engagement. Links below👇
3
6
1
38
6,407
Should protocols get fuzz testing before or after audits/competitions? TLDR at the bottom 👇 Completing the fuzzing suite prior to a audit/competition significantly improves its overall quality. This is because it allows SRs to integrate their own invariants into an existing, top-tier suite, and also it eliminates the need to identify vulnerabilities associated with pre-existing invariants. Consequently, SRs can dedicate more time to uncovering otherwise overlooked vulnerabilities. Even after audits or competitions, fuzz testing remains highly beneficial for a protocol. This is because the issues identified through fuzzing often differ from those found in manual analysis, which will uncover unique vulnerabilities. A recent case study illustrating the benefit of fuzz testing before manual reviews is the recent eBTC protocol competition. Here, the presence of a pre-established fuzzing suite enabled participants to build upon it, and also saving a lot of time for SRs by not needing to verify protocol invariants, leading to a much improved quality of the competition. Link: code4rena.com/audits/2023-10… TLDR: Protocols should always get fuzz testing, since it often finds different issues compared to manual analysis, and it is more beneficial to get it prior to audits.
3
14
1,082
Perimeter retweeted
This made my day 🙂 Another happy customer at @perimeter_sec.
1
1
13
990
Perimeter retweeted
One of my latest fuzzing campaigns is now public. This was a 4 week timeboxed fuzzing engagement for Drips. The first half as part of a security review for @SpearbitDAO, the second part for @perimeter_sec. In the end, zero issues were found. A first for me, and a huge accomplishment by the @dripsnetwork team. Many thanks to @code_sandwich for his feedback and support during the project! Links below.
2
9
62
6,202