DSS 2026 — Auditing Is Not Bug Hunting: Securing Systems Through Risk Management 0xa5df Source video: https://www.youtube.com/watch?v=v_HkiwmN2Zw Automatic captions — not human-reviewed. Complete spoken-audio coverage is unverified. Source language tag: en (English). Source export SHA256: 5f7de7db1045b5f5d29fb02bb773434109e089f64ac3d4ee4338b148488dd087 [0:01] Hi, good afternoon. [0:04] Um, I'm going to talk today about our [0:06] responsibility as auditors and how it [0:09] goes beyond just finding bugs. [0:12] Um, because when I started working at at [0:15] doing the bug hunting and participating [0:17] in contest, my goal was to find as many [0:21] bugs as I can in order to maximize my [0:23] payout. But I didn't have a [0:25] responsibility for the system as a [0:26] whole. If I lost if I missed a bug, it [0:29] was a lost opportunity, not something [0:32] that that I should be too much concerned [0:33] about. [0:35] And then I joined Storra and I started [0:38] to do client audits. And now I started [0:40] to have a responsibility for the system [0:41] as a whole. I had to make sure that I [0:44] don't miss any bugs and especially those [0:47] who lead to [0:50] um who lead to, you know, losses of [0:53] millions of dollars. [0:57] So how did that change my work? In two [1:00] ways. First one is prioritizing. Now in [1:03] bug bounties and contests as well, you [1:05] have to prioritize by severity in order [1:08] to maximize your payout. [1:10] But I would argue that this uh severity [1:13] is often subjective or theoretical and [1:15] doesn't actually represent the actual [1:18] potential damage uh of each bug. [1:22] And when it comes to um actually [1:26] protecting the system, you have to be [1:28] dead practical about the potential [1:30] damage of each bug. [1:33] So usually I classify this into like [1:35] minor issues which are not that [1:36] significant, then significant issues [1:38] which you don't want to miss. But then [1:40] there are those uh issues that make it [1:42] to the news with millions of dollars [1:44] stolen. And this is the kind of stuff [1:46] which you really really don't want to [1:47] miss. [1:50] Now the second way is [1:54] by [1:56] realizing that there are some ways to [1:57] protect the system beyond just finding [1:59] bugs and this is by introducing new [2:01] hardening mechanism which would protect [2:04] the system against some threats which [2:06] are not due to onchain bugs. So a recent [2:08] example is that Hedron um protocol which [2:13] I audited a few months ago [2:15] where [2:17] they have a prop AMM. Now propriety AMM [2:20] unlike a uh decentralized AMM where the [2:23] price is determined by an algorithm. [2:25] Here the price is determined by a single [2:28] owner who owns the liquidity in the pool [2:30] and also updates the price through an [2:33] oracle service. Now this Oracle service [2:37] is um you know it has up to like million [2:41] updates per day. So you can't have like [2:43] a human approving each update with the [2:45] cold wallet. So it's a hot wallet which [2:47] is stored on some server and this is [2:49] obviously has a potential of getting [2:52] compromised. [2:54] Now what happens if it's compromised is [2:57] an attacker can basically steal all of [3:00] the funds in the pool. So they can [3:02] basically set a price so that they can [3:03] give a very fraction of uh soul for [3:06] example for the entire USDC and then [3:08] they give a fraction of use set a price [3:10] to the other way and give a fraction of [3:12] USDC for the remaining soul in the pool [3:15] and this way in a single transaction you [3:17] can just empty the pool. [3:21] So what kind of mechanism did we [3:23] introduce in order to defend against [3:25] this kind of attack is the draw down [3:28] guard. It's basically [3:30] uh a mechanism which monitors that value [3:34] of the liquidity of the pool and if it [3:36] detects that the value is decreasing too [3:38] fast then it pauses the pool and doesn't [3:40] allow any further swap [3:43] because a slow decrease can often happen [3:46] naturally due to price changes. Um, but [3:49] if we see that it's just moving too [3:51] fast, like for example, 2% in 30 [3:54] seconds, then we should pause the pool [3:56] and don't allow any further swap. [4:00] Now, this slows down the attacker. It [4:02] doesn't allow them to empty the pool in [4:04] a single transaction, [4:06] but it still allows them to do a slow [4:09] drain over a few hours. Even with this [4:11] mechanism, they can steal significant [4:14] amount of the pool. So the second part [4:17] of this mechanism is to monitor the pool [4:20] through an external service. Uh a pa a a [4:24] bot which has the pause a role which can [4:26] only pause the the pool and not unpause [4:29] it and in case it detects any invalid [4:31] prices then it pauses the pool and the [4:35] attacker can't um swap anymore. So those [4:38] two mechanism together should [4:40] significantly reduce the total amount of [4:43] funds that an attacker can steal in case [4:45] the uh oracle ro is compromised. [4:50] Uh because you have only a few seconds [4:52] before before the puzzle bot kicks in [4:54] and the amount you can steal is should [4:56] be you know depending on the [4:58] configuration only a few percent um [5:02] before the draw down guard kicks in. [5:06] So yeah, so basically this is an example [5:08] of how we can this is not a bug in the [5:10] system. This is an additional mechanism [5:13] uh which we give to the client to [5:16] protect their system. And basically yeah [5:19] this is this is the reason why the [5:20] clients come to us. They come to us not [5:22] just to find bugs but to protect the [5:24] system and make sure they don't make it [5:26] to the news with millions of dollars [5:27] stolen. So this is what we should [5:29] deliver to them. Thank you.