A credential stuffing attack at the ALB is aLayer 7problem and is best mitigated withAWS WAF. The attacker is distributed across many IPs, so blocking by IP in a security group (Option B) is ineffective and operationally heavy. A CloudWatch alarm (Option A) only alerts; it does not block or mitigate requests.

Because the malicious traffic uses a distinctive, knownUser-Agentstring associated with a mobile device emulator, AWS WAF can quickly reduce the attack by inspecting the User-Agent header and blocking matching requests. This approach is targeted: it blocks the identified automated attack pattern while allowing legitimate users who do not present that emulator User-Agent to continue logging in. The WAF rule can be deployed immediately on the existing ALB-associated web ACL and can be further refined (for example, applied only to /login paths, combined with rate-based rules, or integrated with Bot Control) to minimize false positives.

Option D is risky because ''allow only legitimate user agents'' is brittle: user agents are diverse and change frequently, and a strict allow-list can accidentally block real users. Therefore, a WAF custom block rule for the known malicious User-Agent string is the correct solution.