The digital trading platform Binance explained that the recent breach, which resulted in the seizure of assets worth $6.34 billion, was the result of a complex campaign lasting several months that targeted the computers of three of its wallet infrastructure engineers. The campaign included a breach via a fake job interview and a booby-trapped coding assignment, in addition to exploiting an internal AI tool.
According to the platform's preliminary report, authorities and response teams have managed to freeze and recover approximately $2.85 billion of the stolen assets so far, while the US Federal Bureau of Investigation (FBI) attributed the attack to the cyber infiltration group "TraderTraitor" linked to North Korea.
Details of the attackers' infiltration via social engineering
The infiltration process began in June when a Wallet DevOps engineer attempted to download an AI-based programming tool via a search engine, leading to the installation of a malicious modified version of the Homebrew package. Although the platform's cybersecurity team detected the threat, replaced the engineer's device, and reset his credentials, the attackers managed to collect valuable information about the nature of his work and his team.
In September, attackers targeted a second engineer through a recruitment scam, where two interviews were conducted; the first with a fake HR representative who provided social media accounts verified with personal photos to gain trust, and the second with a person claiming to be the CEO of a fake company. The interview ended with a request to download a "coding assignment" to test his skills, which contained a malicious code repository that executed immediately upon setup. The malware replaced a key component in the development tools and opened a communication channel to the attackers' command server without being detected by Endpoint Detection and Response (EDR) systems, as it relied on a zero-day vulnerability.
Exploiting the AI tool and protocol vulnerabilities
On October 20, the attackers used the infected computer in addition to an internal AI tool to access the system of the first engineer and the system of a third engineer in the wallet team. On October 29, they accessed the third engineer's computer and planted a container environment using the first engineer's credentials, then ran a malicious script to transfer assets.
The attackers exploited unauthorized access to the hot wallet signing service to transfer currencies such as XRP, SOL, OP, and BTC. They also exploited the EIP-7702 mechanism on the Ethereum network to grant a specific smart contract partial execution permissions on the hot wallet. Since the wallet was protected by Multi-Party Computation (MPC) technology, there were no single keys to immediately revoke authorization, and transfers continued until authorization was revoked after midnight.
Why does this news matter to the tech and cyber sectors?
This incident reveals a qualitative shift in social engineering methods; the threat is no longer limited to traditional phishing messages but has extended to include fully planned job interviews and the booby-trapping of daily development tools. It also highlights the dual challenge resulting from integrating AI tools within corporate environments without strict permission constraints, as they can be exploited as a bridge to access sensitive systems and perform lateral movement between authorized devices.
Recovery and security measures taken
Binance confirmed the reconstruction of the wallet system and the affected AI tool from scratch, the replacement of hardware, and the revocation of all previous tokens and credentials. It also engaged independent security firms to review the incident and announced the establishment of a research fund specialized in AI security within blockchain environments.
For his part, Co-CEO Richard Teng stated that 98% of the affected assets have been compensated and their stability restored, and withdrawal operations have been made gradually available to users, in parallel with the deployment of Zero-Knowledge Proofs to confirm the full coverage of the platform's reserves.