01 · Detects
Adaptive Detection
→ MORE
No static filter. No manual rule tuning.
ICN learns continuously from delivery outcomes, sender behaviour, and network feedback. Drift and anomalies are detected before they cause damage.
An adaptive delivery decision system that controls, optimises, and stabilises email flows in real time — fully automated, with no manual intervention.
Two roles, one pipeline. Optimisation and security are computed simultaneously — not processed sequentially.
Optimisation and security are computed simultaneously — not one after the other.
The system adapts on an ongoing basis — not through manual rules, but through observed behaviour. Every signal feeds back into the decision process.
Every outbound email passes through all stages sequentially. No additional messaging layers, no network hops, no unnecessary latency.
No single signal triggers a block. The pipeline requires consistent negative signals across multiple independent algorithms — an attacker would need to simultaneously fool the reputation algorithm, the anomaly detector, and the duplicate checker.
Security checks take precedence over performance optimisations. The premium hourly limit is domain-based — multiple sender addresses sharing a domain share a single quota.
Score too high or rate limit exceeded. No retry — full logging for analysis.
Normal outbound mail without premium status. Classic MX path with conservative limits.
Consumer domains (Gmail, Yahoo, Microsoft). Reserved for demonstrably clean senders with stable reputation.
Bulk mailings with LSTO header. Gmail- and Yahoo-compliant for high-volume legitimate campaigns.
Built for environments where delivery is not binary — where reputation, context, and behaviour all count.
ICN does not replace Rspamd — it builds on Rspamd results and condenses them through behavioural, historical, and contextual signals into robust delivery decisions.
| Feature | Classic MTA + Rspamd | ICN |
|---|---|---|
| Reputation learning | None — static rule configuration | Online learning with exponential decay |
| Anomaly detection | Absolute thresholds | Behaviour-based, time-of-day adaptive |
| Bounce processing | Manual or rudimentary | RFC 3464, fully automated, idempotent |
| False-positive protection | No systematic mechanism | Legitimacy detection, multi-signal voting |
| Authentication chain | DKIM/SPF/DMARC passive | ARC signing per RFC 8617, active |
| Routing granularity | One or two paths | Four delivery zones, domain-granular |
A 45-minute architecture session with a delivery engineer. We analyse your current outbound topology and show exactly where ICN would integrate. Outcome: a technical assessment, not a sales deck.