TL;DR: Evasive attacker behaviour still slips past signatures and anomaly models, Corelight reports, because one is reactive and the other needs time to learn a baseline. Detection programmes need technique-level logic that analysts can inspect and trust.
Editorial analysis by NHI Mgmt Group, based on content published by Corelight: “The missing layer in modern detection architecture”.
By the numbers:
- Anomaly detection often needs up to 31 days to build a reliable baseline, according to Corelight.
Key questions
Q: Why do evasive attackers defeat common detection models so often?
A: They change the wrapper, not the method.
Q: What should organisations measure to know whether behavioural detection is working?
A: They should measure whether the platform flags abnormal sequences across email, identity, and SaaS activity before the attacker completes a fraudulent action.
Q: What breaks when signature and anomaly detection are the only controls in place?
A: Teams lose the middle layer that identifies attacker tradecraft when the adversary is not using a known fingerprint and is not noisy enough to stand out statistically.
Practitioner guidance
- Define ATT&CK-mapped detection coverage Inventory which attacker techniques your current signatures, anomaly models, and IOC feeds actually cover, then identify the gaps where technique-based detections are missing.
- Prioritise explainable detections for high-risk techniques Make credential access, lateral movement, and command-and-control the first techniques you cover with logic that analysts can inspect and test.
- Validate behavioural logic before relying on it Run detections through controlled testing and review the underlying conditions, not just the alert outcome, so false confidence does not become operational debt.
Bottom line: The article’s central argument is that neither signatures nor anomaly detection fully closes the gap against evasive adversaries operating under normal-looking traffic patterns.
Glossary terms: Ttp-Based Detection, Detection as code, Anomaly Detection, Kill Chain
Explore further
View Full Forum → | NHI Foundation Course → | Our Services → | Read the full analysis →
Behavioural detection becomes a governance problem when vendors hide the logic. If a detection cannot be inspected, tuned, or mapped to a specific attacker technique, analysts inherit alerts without assurance. That weakens trust in the control itself, especially where response decisions depend on confidence rather than volume. The more evasive the adversary, the more important it becomes that detection logic is explainable and testable.
A question worth separating out:
Q: Should SOC teams prioritise ATT&CK-mapped detections over broader anomaly models?
A: They should treat them as complementary, not competing. ATT&CK-mapped detections are better for precise, high-confidence identification of known techniques, while anomaly models help surface unknown or unusual behaviour. The right sequencing is to use technique-based coverage for the highest-risk paths and anomaly detection where behavioural novelty matters most.
👉 Read our full editorial: TTP-based behavioral detection fills the gap between signatures and anomalies