In a recent cybersecurity incident, Google’s response team in the United States investigated a severe...

An HTTP GET POST attack is the most likely answer because the key indicator in the scenario is the measurement unit and behavior: 398 million requests per second. In CEH-aligned DoS and DDoS coverage, volumetric network-layer attacks like SYN floods are typically discussed in packets per second or bits per second, and they primarily exhaust connection tables or bandwidth at Layer 3 and Layer 4. By contrast, an HTTP flood targets Layer 7 and is commonly measured in requests per second because the attacker is sending large volumes of seemingly valid web requests. This overwhelms web servers, application stacks, and upstream components such as reverse proxies, load balancers, and API gateways, leading to degraded performance and dropped sessions, exactly as described when “active connections drop unexpectedly.”

The mention of “customer-facing platforms” and “numerous compromised devices” also matches typical botnet-driven application-layer flooding, where many distributed sources generate massive HTTP request rates to evade simple IP-based blocking. A TCP SACK panic attack is a vulnerability-triggered crash condition related to TCP handling, not a request-rate spike scenario. An RST attack involves sending TCP reset packets to tear down sessions, which would not normally present as an extreme requests-per-second event in server logs.

Therefore, the evidence most strongly supports an application-layer DDoS: an HTTP GET POST flood designed to exhaust server and application resources and cause intermittent service disruption.