DEV Community

Saurabh Kumar
Saurabh Kumar

Posted on

Why CIDS Looks at Behavior, Not Just Individual Requests

Security systems are very good at answering one question:

“Is this request suspicious?”

But sometimes that is not enough.

An attacker does not always reveal their intent through a single request.

The intent can emerge through a sequence of actions.

That is one of the main ideas behind CyberMoranda CIDS.

A Single Request Can Be Misleading

Consider a normal-looking application request:

*GET /admin
*

By itself, this may not be enough to classify the activity as malicious.

Now consider the sequence:

GET /login

Multiple failed authentication attempts

GET /robots.txt

GET /admin

GET /.env

Multiple endpoint probes

Abnormally high request frequency

The individual requests are not the complete story.

The sequence is the story.

CIDS is being designed to preserve and analyze that context.

From Events to Behavior

A traditional event-based approach might look like:

Request

Rule Match

Allow / Block

CIDS is being designed around:

Request

Session Context

Previous Activity

Behavioral Signals

Risk Evaluation

Decision

This means that the system doesn't have to make every security decision from a single isolated event.

What Is a Behavioral Signal?

A behavioral signal is an observable characteristic of activity that contributes evidence to a security decision.

For example:

Authentication failures
Endpoint enumeration
Request frequency
Restricted path access
Session history
Fingerprint changes
Unusual request sequences

Each signal can contribute different evidence.

The important part is that the system should be able to explain the evidence.

Instead of:

Threat detected.
Risk = 87

CIDS should eventually be able to provide something closer to:

Risk Score: 87

Contributing signals:

  • Repeated authentication failures
  • Restricted endpoint probing
  • Unusual request frequency
  • Suspicious request sequence

Confidence: High
Decision: Contain

That makes the decision much easier for a security analyst to understand.

Context Changes the Meaning of an Event

Imagine an authenticated employee accessing:

/api/profile

That may be completely normal.

The same identity suddenly starts doing:

*/.env
/admin
/config
/internal/

**
within a short period of time.

The important question is no longer simply:

“Is this user authenticated?”

The better question becomes:

“Is this behavior consistent with the context of this session?”

This is where behavioral security becomes interesting.

Risk Should Be Accumulated From Evidence

CIDS is being designed around an incremental risk model rather than a single binary verdict.

Conceptually:

┌─ Signal A

Session ────────────┼─ Signal B

├─ Signal C

└─ Historical Context

Risk Evaluation

Policy Decision

The exact scoring model still needs to be tested and calibrated.

That is important.

A security system should not produce impressive-looking numbers without evidence behind them.

Detection and Response Should Be Separate

Another principle I'm following is to keep detection separate from response.

*Behavior

Detection

Risk

Policy

Response
*

Why?

Because detecting suspicious behavior and deciding what to do about it are two different problems.

One environment might want:

**High Risk → Monitor

Another might want:

High Risk → Rate Limit

Another might want:

Critical → Contain**

Separating these layers makes the system more adaptable.

The Goal Isn't More Blocking

This is probably the most important part.

CIDS isn't being designed with the philosophy:

“Block everything suspicious.”

The objective is:

Understand enough context to choose an appropriate defensive response.

That response could be monitoring.

It could be rate limiting.

It could be containment.

And in controlled circumstances, it could involve deception.

The system should remain defensive and should never retaliate against an attacker.

Where This Goes Next

Behavioral detection is only one layer.

The larger CIDS architecture is moving toward combining:

*Network / Application Events
+
Host Telemetry
+
Fingerprint Signals
+
Session Context
+
Threat Intelligence
+
Behavioral Analysis

Security Signals

Risk Engine

Policy Engine

Adaptive Response
*

The difficult part isn't drawing this architecture.

The difficult part is building it, testing it, measuring it, and proving that it works without creating unacceptable false positives or performance problems.

That's the part I'm working toward.

CIDS is not finished.

I'm building the foundation first.

Observe the behavior.
Understand the context.
Then act.

*CyberMoranda CIDS — Think Before You Act.
*

cybersecurity

threatdetection

security

rust

opensource

Top comments (0)