North Korea's Cryptocurrency Theft Tactics Are Evolving, with DeFi Being a Frequent Target

Less than three weeks after hackers linked to North Korea used social engineering to breach the cryptocurrency trading firm Drift, another major exploit has been carried out, this time targeting Kelp, a restaking protocol connected to LayerZero's cross-chain infrastructure. This suggests an evolution in the tactics of North Korea-linked hackers, who are no longer just looking for vulnerabilities or stolen credentials but are instead exploiting the fundamental assumptions built into decentralized systems. The combined incidents of the Drift and Kelp exploits, which together resulted in the theft of over $500 million in just over two weeks, point to a more organized effort by North Korea to hijack funds from the cryptocurrency sector. According to Alexander Urbelis, Chief Information Security Officer and General Counsel at ENS Labs, 'This is not a series of incidents; it is a cadence. You cannot patch your way out of a procurement schedule.' The Kelp exploit did not involve breaking encryption or cracking keys; instead, the attackers manipulated the data feeding into the system, forcing it to rely on compromised inputs and approve transactions that never actually occurred. As Urbelis noted, 'The security failure is simple: a signed lie is still a lie. Signatures guarantee authorship; they do not guarantee truth.' This highlights that the system checked who sent the message, not whether the message itself was correct, making this exploit more about how the system was set up rather than a clever new hack. David Schwed, COO of blockchain security firm SVRN, agreed, stating, 'This attack wasn’t about breaking cryptography. It was about exploiting how the system was set up.' A key issue was a configuration choice, with Kelp relying on a single verifier to approve cross-chain messages, which, although faster and simpler to set up, removes a critical safety layer. In response, LayerZero has recommended using multiple independent verifiers to approve transactions, akin to requiring multiple signatures on a bank transfer. However, some have pushed back on this, arguing that LayerZero’s default setup was to have a single verifier. Schwed emphasized, 'If you’ve identified a configuration as unsafe, don’t ship it as an option. Security that depends on everyone reading the docs and getting it right is not realistic.' The fallout from the exploit has not been contained to Kelp, as its assets are used across multiple platforms, causing problems to spread. As Schwed explained, 'These assets are a chain of IOUs. And the chain is only as strong as the controls on each link.' When one link breaks, others are affected, and in this case, lending platforms like Aave that accepted the impacted assets as collateral are now dealing with losses, turning a single exploit into a wider stress event. The attack also exposes a gap between the marketing of decentralization and its actual implementation. As Schwed pointed out, 'A single verifier is not decentralized. It’s a centralized decentralized verifier.' Urbelis broadened this perspective, stating, 'Decentralization is not a property a system has. It is a series of choices. And the stack is only as strong as its most centralized layer.' In practice, this means even systems that appear decentralized can have weak points, especially in less visible layers like data providers or infrastructure, which are increasingly the focus of attackers. The recent targeting by the Lazarus group of cross-chain and restaking infrastructure, critical but complex layers that move assets between systems or allow them to be reused, may explain this shift. These layers are not only hard to monitor and easy to misconfigure but also hold large amounts of value, making them attractive targets. As Lazarus adapts, the biggest risk may not be unknown vulnerabilities but known ones that are not fully addressed. The Kelp exploit did not introduce a new kind of weakness; it showed how exposed the ecosystem remains to familiar ones, especially when security is treated as a recommendation rather than a requirement. And as attackers move faster, this gap is becoming both easier to exploit and far more expensive to ignore.