update

README.md

@@ -210,4 +210,164 @@ tags:
 - cybersecurity
 size_categories:
 - 1M<n<10M
+---
+# CICFlow Binary Intrusion Detection Dataset
+
+## Overview
+
+This dataset provides a **binary network intrusion detection (IDS) benchmark** derived from **CICFlowMeter flow-level features**.
+
+The task is to determine whether a given network flow is **benign** or **malicious**, making this dataset suitable for:
+
+* Practical IDS research
+* Rule-based and hybrid ML systems
+* High-imbalance classification experiments
+* Explainable security modeling
+
+---
+
+## Task
+
+**Binary classification**
+
+> *Given a network flow represented by CICFlowMeter features, predict whether it corresponds to an attack.*
+
+---
+
+## Labels
+
+| Label ID | Name | Description |
+| -------: | ------ | ----------------------------------------------- |
+| 0 | benign | Normal network traffic |
+| 1 | attack | Any malicious traffic (all attack types merged) |
+
+### Label construction
+
+The binary label was derived by collapsing a multiclass IDS taxonomy:
+
+* `benign` → `0`
+* all attack categories → `1`
+
+No other transformations were applied to the labels.
+
+---
+
+## Dataset Structure
+
+```text
+DatasetDict({
+train,
+validation,
+test
+})
+```
+
+Each split contains records with the following schema:
+
+```text
+flow_id: string
+features: dict[str, float]
+semantic_flags: dict[str, int]
+label: ClassLabel (benign / attack)
+```
+
+---
+
+## Feature Description
+
+### 1. Raw Numeric Features (`features`)
+
+The `features` field contains **flow-level statistics extracted by CICFlowMeter**.
+
+These include:
+
+#### Traffic volume & direction
+
+* `total_fwd_packets`, `total_bwd_packets`
+* `total_length_of_fwd_packets`
+* `total_length_of_bwd_packets`
+* `down_up_ratio`
+
+#### Packet size statistics
+
+* `packet_length_min`, `packet_length_max`
+* `packet_length_mean`, `packet_length_std`, `packet_length_variance`
+* Forward and backward packet length metrics
+
+#### Timing & inter-arrival times
+
+* `flow_duration`
+* `flow_iat_mean`, `flow_iat_std`, `flow_iat_max`, `flow_iat_min`
+* Forward and backward IAT statistics
+* `active_*`, `idle_*` features
+
+#### Rate-based features
+
+* `flow_bytes_per_s`
+* `flow_packets_per_s`
+* `fwd_packets_per_s`, `bwd_packets_per_s`
+
+#### TCP flag counters
+
+* `syn_flag_count`, `ack_flag_count`, `rst_flag_count`
+* `fin_flag_count`, `psh_flag_count`, `urg_flag_count`
+* `cwr_flag_count`, `ece_flag_count`
+
+#### Bulk and subflow statistics
+
+* `*_bulk_*` features
+* `subflow_fwd_*`, `subflow_bwd_*`
+
+> **Note:** Some CICFlowMeter features are conditionally emitted.
+> Missing or undefined numeric values were **filled with `0.0`**, which semantically indicates *absence of that behavior*.
+
+---
+
+### 2. Semantic Flags (`semantic_flags`)
+
+Each flow includes deterministic, interpretable indicators derived from numeric features:
+
+| Flag | Meaning |
+| --------------------- | --------------------------------------------------------- |
+| `high_packet_rate` | Extremely high packets per second |
+| `one_way_traffic` | Forward-only traffic (no backward response) |
+| `syn_ack_imbalance` | Large SYN/ACK imbalance |
+| `uniform_packet_size` | Low packet size variance |
+| `long_idle_c2` | Long idle periods (possible command-and-control behavior) |
+
+These flags are:
+
+* Deterministic
+* Stable across dataset versions
+* Suitable for rule-based IDS and hybrid systems
+
+---
+
+## Preprocessing Summary
+
+The dataset was generated using a **fully deterministic preprocessing pipeline**:
+
+* Column name normalization
+* Conversion of `NaN` / `±Infinity` → `0.0`
+* Explicit label normalization
+* No row removal due to missing numeric values
+* Stratified train/validation/test splits
+* Explicit Hugging Face feature schemas
+
+No rebalancing, oversampling, or augmentation was applied.
+
+---
+
+## Class Imbalance
+
+The dataset is **highly imbalanced**, with benign traffic dominating.
+
+This reflects **real-world network conditions** and is intentional.
+
+Users are encouraged to:
+
+* Use **precision, recall, F1, PR-AUC**
+* Apply **class weighting** or cost-sensitive learning
+* Avoid relying on accuracy alone
+
 ---