Steakhouse postmortem reveals DNS hijack caused by registrar 2FA bypass

ambcryptoPublicado em 2026-04-10Última atualização em 2026-04-10

Resumo

Steakhouse's postmortem of a 30 March security incident reveals that attackers hijacked its domain through a social engineering attack on its registrar, OVHcloud. The attacker impersonated the account owner, convinced support to disable hardware-based two-factor authentication, and took full control of the account. This allowed them to redirect DNS to a phishing site with a wallet drainer for about four hours. No user funds were lost, as on-chain systems remained secure, and wallet protections quickly detected the fake site. The breach underscores the risk of off-chain infrastructure vulnerabilities and over-reliance on a single registrar. Steakhouse has since migrated registrars, enhanced DNS monitoring, and implemented stricter domain security controls.

A postmortem from Steakhouse has shed new light on a 30 March security incident. Attackers briefly hijacked its domain to serve a phishing site, exposing a critical weakness in off-chain infrastructure rather than on-chain systems.

The team confirmed that the attack stemmed from a successful social engineering attempt targeting its domain registrar, OVHcloud. This allowed the attacker to bypass two-factor authentication and take control of DNS records.

Social engineering led to full account takeover

According to the report, the attacker contacted the registrar’s support desk, impersonated the account owner, and convinced a support agent to remove hardware-based two-factor authentication.

Once access was granted, the attacker rapidly executed a series of automated actions. This included deleting existing security credentials, enrolling new authentication devices, and redirecting DNS records to infrastructure under their control.

This enabled the deployment of a cloned Steakhouse website embedded with a wallet drainer, which remained intermittently accessible for roughly four hours.

Phishing site active, but funds remained safe

Despite the severity of the breach, Steakhouse stated that no user funds were lost and no malicious transactions were confirmed.

The compromise was limited to the domain layer. On-chain vaults and smart contracts, which operate independently of the frontend, were not affected. The protocol emphasized that it holds no admin keys that could access user deposits.

Browser wallet protections from providers such as MetaMask and Phantom quickly flagged the phishing site, while the team issued a public warning within 30 minutes of detecting the incident.

Postmortem highlights vendor risk and single points of failure

The report points to a key failure in Steakhouse’s security assumptions: reliance on a single registrar whose support processes could override hardware-based protections.

The ability to disable two-factor authentication via a phone call, without robust out-of-band verification, effectively turned a credential leak into a full account takeover.

Steakhouse acknowledged that it had not adequately assessed this risk, describing the registrar as a “single point of failure” in its infrastructure.

Off-chain vulnerabilities remain a weak link

The incident underscores a broader issue in crypto security — that strong on-chain protections do not eliminate risks in surrounding infrastructure.

While smart contracts and vaults remained secure, control over DNS allowed the attacker to target users through phishing, a method increasingly common in the ecosystem.

The attack also involved tools consistent with “drainer-as-a-service” operations, highlighting how attackers continue to combine social engineering with ready-made exploit kits.

Security upgrades and next steps

Following the incident, Steakhouse has migrated to a more secure registrar. It implemented continuous DNS monitoring, rotated credentials, and launched a broader review of vendor security practices.

The team also introduced stricter controls for domain management, including hardware key enforcement and registrar-level locks.


Final Summary

  • Steakhouse’s postmortem reveals that a registrar-level 2FA bypass enabled a DNS hijack, exposing users to phishing despite secure on-chain systems.
  • The incident highlights how off-chain infrastructure and vendor security remain critical vulnerabilities in crypto ecosystems.

Perguntas relacionadas

QWhat was the root cause of the security incident at Steakhouse on March 30th?

AThe root cause was a successful social engineering attack targeting their domain registrar, OVHcloud, which allowed the attacker to bypass two-factor authentication and take control of the DNS records.

QHow did the attacker manage to bypass the two-factor authentication on the registrar account?

AThe attacker impersonated the account owner, contacted the registrar's support desk, and convinced a support agent to remove the hardware-based two-factor authentication protection.

QWere any user funds lost as a result of this DNS hijacking and phishing attack?

ANo, Steakhouse confirmed that no user funds were lost and no malicious transactions were confirmed. The on-chain vaults and smart contracts were not compromised.

QWhat key security failure did the postmortem report identify in Steakhouse's infrastructure?

AThe report identified the reliance on a single registrar, whose support processes could override hardware-based protections, as a critical 'single point of failure' that was not adequately assessed.

QWhat security measures did Steakhouse implement after the incident to prevent future attacks?

ASteakhouse migrated to a more secure registrar, implemented continuous DNS monitoring, rotated credentials, enforced stricter domain management controls (like hardware keys), and launched a broader review of vendor security practices.

Leituras Relacionadas

OpenAI Post-Training Engineer Weng Jiayi Proposes a New Paradigm Hypothesis for Agentic AI

OpenAI engineer Weng Jiayi's "Heuristic Learning" experiments propose a new paradigm for Agentic AI, suggesting that intelligent agents can improve not just by training neural networks, but also by autonomously writing and refining code based on environmental feedback. In the experiment, a coding agent (powered by Codex) was tasked with developing and maintaining a programmatic strategy for the Atari game Breakout. Starting from a basic prompt, the agent iteratively wrote code, ran the game, analyzed logs and video replays to identify failures, and then modified the code. Through this engineering loop of "code-run-debug-update," it evolved a pure Python heuristic strategy that achieved a perfect score of 864 in Breakout and performed competitively with deep reinforcement learning (RL) algorithms in MuJoCo control tasks like Ant and HalfCheetah. This approach, termed Heuristic Learning (HL), contrasts with Deep RL. In HL, experience is captured in readable, modifiable code, tests, logs, and configurations—a software system—rather than being encoded solely into opaque neural network weights. This offers potential advantages in explainability, auditability for safety-critical applications, easier integration of regression tests to combat catastrophic forgetting, and more efficient sample use in early learning stages, as demonstrated in broader tests on 57 Atari games. However, the blog acknowledges clear limitations. Programmatic strategies struggle with tasks requiring long-horizon planning or complex perception (e.g., Montezuma's Revenge), areas where neural networks excel. The future vision is a hybrid architecture: specialized neural networks for fast perception (System 1), HL systems for rules, safety, and local recovery (also System 1), and LLM agents providing high-level feedback and learning from the HL system's data (System 2). The core proposition is that in the era of capable coding agents, a significant portion of an AI's learned experience could be maintained as an auditable, evolving software system.

marsbitHá 50m

OpenAI Post-Training Engineer Weng Jiayi Proposes a New Paradigm Hypothesis for Agentic AI

marsbitHá 50m

Your Claude Will Dream Tonight, Don't Disturb It

This article explores the recent phenomenon of AI companies increasingly using anthropomorphic language—like "thinking," "memory," "hallucination," and now "dreaming"—to describe machine learning processes. Focusing on Anthropic's newly announced "Dreaming" feature for its Claude Agent platform, the piece explains that this function is essentially an automated, offline batch processing of an agent's operational logs. It analyzes past task sessions to identify patterns, optimize future actions, and consolidate learnings into a persistent memory system, akin to a form of reinforcement learning and self-correction. The article draws parallels to similar features in other AI agent systems like Hermes Agent and OpenClaw, which also implement mechanisms for reviewing historical data, extracting reusable "skills," and strengthening long-term memory. It notes a key difference from human dreaming: these AI "dreams" still consume computational resources and user tokens. Further context is provided by discussing the technical challenges of managing AI "memory" or context, highlighting the computational expense of large context windows and innovations like Subquadratic's new model claiming drastically longer contexts. The core critique argues that this strategic use of human-centric vocabulary does more than market products; it subtly reshapes user perception. By framing algorithms with terms associated with consciousness, companies blur the line between tool and autonomous entity. This linguistic shift can influence user expectations, tolerance for errors, and even perceptions of responsibility when systems fail, potentially diverting scrutiny from the companies and engineers behind the technology. The article concludes by speculating that terms like "daydreaming" for predictive task simulation might be next, continuing this trend of embedding the idea of an "inner life" into computational processes.

marsbitHá 52m

Your Claude Will Dream Tonight, Don't Disturb It

marsbitHá 52m

Trading

Spot
Futuros
活动图片