Design Email Security Gateway - System Design | WinJob

Problem Statement and Email Gateway Placement

SMTP sessions arrive at regional MTA pools. Each message is normalized to a canonical envelope (headers, MIME tree hash, attachment manifest). Cheap checks run first: RBL, SPF/DKIM/DMARC alignment, known-good allowlists. Suspicious MIME branches to sandbox queues; URL links are rewritten for time-of-click analysis.

Design anchors (1)

MX records point to your gateway cluster before mail reaches Google Workspace or Exchange.
Separate inbound inspection (spam, phishing, malware) from outbound DLP (data loss, encryption policy).
Quarantine is a first-class workflow—users and SOC must release with audit, not shadow-delete.

Failure drills

If AV engine farm is degraded, degrade to single-engine scan with longer queue delay—never pass unscanned executables. If DMARC strict policy blocks legitimate forwarders, route to admin review bucket with 72h TTL. If click-time proxy is down, strip rewrite and block external links in HTML bodies for high-risk tenants.

Cost and capacity

At 8M messages/day inbound, CPU is dominated by MIME recursion and sandbox VM minutes—not SMTP handshake.

Signal	Target
Peak msg/s	120 /s
Avg size	85 KB
Quarantine	2.1 %
FP budget	0.05 %

javaOne Dark Pro

1 public record MailVerdict(String tenantId, String policyVersion, String disposition, int ruleId) {
2   public boolean requiresQuarantine() { return "quarantine".equals(disposition); }
3 }

pythonOne Dark Pro

1 def dmarc_aligned(spf: str, dkim: str, policy: str) -> bool:
2     return spf == "pass" and dkim == "pass" and policy != "reject"

typescriptOne Dark Pro

1 export interface NormalizedEnvelope {
2   messageId: string;
3   fromDomain: string;
4   mimeParts: number;
5 }
6 export function cacheKey(tenantId: string, env: NormalizedEnvelope): string {
7   return `${tenantId}:${env.messageId}:${env.mimeParts}`;
8 }

Why interviewers care

Email Security Gateway interviews reward crisp scope, explicit trade-offs, and failure stories—not generic microservice diagrams.

Interview checkpoint

Name one failure story for Problem Statement and Email Gateway Placement that proves you understand real outages, not happy-path diagrams.

Key Highlights

•MX records point to your gateway cluster before mail reaches Google Workspace or
•Metrics: Peak msg/s, Avg size, Quarantine, FP budget
•Separate **inbound inspection** (spam, phishing, malware) fr

1	public record MailVerdict(String tenantId, String policyVersion, String disposition, int ruleId) {
2	public boolean requiresQuarantine() { return "quarantine".equals(disposition); }
3	}

1	def dmarc_aligned(spf: str, dkim: str, policy: str) -> bool:
2	return spf == "pass" and dkim == "pass" and policy != "reject"

1	export interface NormalizedEnvelope {
2	messageId: string;
3	fromDomain: string;
4	mimeParts: number;
5	}
6	export function cacheKey(tenantId: string, env: NormalizedEnvelope): string {
7	return `${tenantId}:${env.messageId}:${env.mimeParts}`;
8	}