DMIT：攻击者正在大规模利用 TCP 443 应用设计缺陷，向 OpenAI 发送超大规模请求

原文：

DMIT proactive security detected that 2% of tenants were deploying potentially risky applications in the DMIT LAX data center. These tenants’ applications sent significant large traffic to China several hours ago.

• DMIT detected abnormal traffic within the network 12 hours ago but took no action due to complexity.


• However, this potential risk factor was officially exploited 8 hours ago.


• DMIT will officially notify affected customers within 12 hours and require them to resolve the issue within 24 hours to avoid service interruption for a long time and abuse penalties.


• The relevant users will receive an email soon, and the VM will be immediately shut down to avoid the transfer being used up. (DMIT will not recover this transfer quota since it was due to the client’s unsafe application.)

This vulnerable application cannot directly generate such a large amount of traffic, even though it is open to the internet. Exploiting this type of vulnerability requires real data and requests; it does not possess amplification characteristics.

DMIT initially speculates that attackers in restricted areas are exploiting this TCP 443 application design flaw on a large scale to make super-large-scale real and accepted requests to OpenAI.

Further details will be officially provided after the issue is resolved.