Attack detection — in practice

Published on 2026-07-19 · 7 min read

What this is about

A layer 7 attack often looks harmless on the network graph because bandwidth stays unremarkable. When in doubt: observe before blocking — a false positive costs more than a scan that got through. Filtering only at the origin means you have already paid for the capacity you meant to protect. Before any rule change, it should be clear what share of real traffic would be affected.

Before any rule change, it should be clear what share of real traffic would be affected. Filtering only at the origin means you have already paid for the capacity you meant to protect. A rule that never fires is more dangerous than no rule — it manufactures a false sense of safety. The threshold that was right yesterday is wrong again after a marketing campaign goes out. Before any rule change, it should be clear what share of real traffic would be affected.

What happens in practice

What matters is not the packet rate but how expensive a single request is for the origin. Before any rule change, it should be clear what share of real traffic would be affected. Filtering only at the origin means you have already paid for the capacity you meant to protect. When in doubt: observe before blocking — a false positive costs more than a scan that got through. A rule that never fires is more dangerous than no rule — it manufactures a false sense of safety. A rule that never fires is more dangerous than no rule — it manufactures a false sense of safety.

What matters is not the packet rate but how expensive a single request is for the origin. A rule that never fires is more dangerous than no rule — it manufactures a false sense of safety. Filtering only at the origin means you have already paid for the capacity you meant to protect. Before any rule change, it should be clear what share of real traffic would be affected. Filtering only at the origin means you have already paid for the capacity you meant to protect. A rule that never fires is more dangerous than no rule — it manufactures a false sense of safety.

What matters is not the packet rate but how expensive a single request is for the origin. Before any rule change, it should be clear what share of real traffic would be affected. A rule that never fires is more dangerous than no rule — it manufactures a false sense of safety. When in doubt: observe before blocking — a false positive costs more than a scan that got through. What matters is not the packet rate but how expensive a single request is for the origin. A layer 7 attack often looks harmless on the network graph because bandwidth stays unremarkable.

Filtering only at the origin means you have already paid for the capacity you meant to protect.
What matters is not the packet rate but how expensive a single request is for the origin.
When in doubt: observe before blocking — a false positive costs more than a scan that got through.

What we recommend

A layer 7 attack often looks harmless on the network graph because bandwidth stays unremarkable. When in doubt: observe before blocking — a false positive costs more than a scan that got through. Before any rule change, it should be clear what share of real traffic would be affected. A rule that never fires is more dangerous than no rule — it manufactures a false sense of safety.

A rule that never fires is more dangerous than no rule — it manufactures a false sense of safety. A layer 7 attack often looks harmless on the network graph because bandwidth stays unremarkable. When in doubt: observe before blocking — a false positive costs more than a scan that got through. Before any rule change, it should be clear what share of real traffic would be affected. Before any rule change, it should be clear what share of real traffic would be affected.

What matters is not the packet rate but how expensive a single request is for the origin.
Before any rule change, it should be clear what share of real traffic would be affected.
Before any rule change, it should be clear what share of real traffic would be affected.

Common mistakes

Before any rule change, it should be clear what share of real traffic would be affected. Filtering only at the origin means you have already paid for the capacity you meant to protect. When in doubt: observe before blocking — a false positive costs more than a scan that got through. A layer 7 attack often looks harmless on the network graph because bandwidth stays unremarkable. A rule that never fires is more dangerous than no rule — it manufactures a false sense of safety.

A rule that never fires is more dangerous than no rule — it manufactures a false sense of safety. The threshold that was right yesterday is wrong again after a marketing campaign goes out. What matters is not the packet rate but how expensive a single request is for the origin. The threshold that was right yesterday is wrong again after a marketing campaign goes out.

What matters is not the packet rate but how expensive a single request is for the origin. A rule that never fires is more dangerous than no rule — it manufactures a false sense of safety. When in doubt: observe before blocking — a false positive costs more than a scan that got through. When in doubt: observe before blocking — a false positive costs more than a scan that got through.

Before any rule change, it should be clear what share of real traffic would be affected.
The threshold that was right yesterday is wrong again after a marketing campaign goes out.
Before any rule change, it should be clear what share of real traffic would be affected.

In summary

A rule that never fires is more dangerous than no rule — it manufactures a false sense of safety. When in doubt: observe before blocking — a false positive costs more than a scan that got through. A layer 7 attack often looks harmless on the network graph because bandwidth stays unremarkable. When in doubt: observe before blocking — a false positive costs more than a scan that got through. When in doubt: observe before blocking — a false positive costs more than a scan that got through.

What matters is not the packet rate but how expensive a single request is for the origin. A layer 7 attack often looks harmless on the network graph because bandwidth stays unremarkable. A rule that never fires is more dangerous than no rule — it manufactures a false sense of safety. The threshold that was right yesterday is wrong again after a marketing campaign goes out. A layer 7 attack often looks harmless on the network graph because bandwidth stays unremarkable.

When in doubt: observe before blocking — a false positive costs more than a scan that got through. Before any rule change, it should be clear what share of real traffic would be affected. Before any rule change, it should be clear what share of real traffic would be affected. When in doubt: observe before blocking — a false positive costs more than a scan that got through.

What matters is not the packet rate but how expensive a single request is for the origin.
What matters is not the packet rate but how expensive a single request is for the origin.
Filtering only at the origin means you have already paid for the capacity you meant to protect.