Ethereum Fraud Detection

<title>Ethereum Fraud Detection — README</title> <style> :root { --bg: #0f1117; --bg2: #161b25; --bg3: #1c2333; --card: #1a2030; --border: rgba(120, 160, 220, 0.12); --border2: rgba(120, 160, 220, 0.25); --blue: #7eb8f7; --teal: #56d4bc; --red: #f87171; --yellow: #fbbf5a; --green: #6ee7b7; --text: #dde3ee; --text2: #8a97b0; --text3: #4e5d75; --mono: 'IBM Plex Mono', monospace; --sans: 'IBM Plex Sans', sans-serif; }

* { box-sizing: border-box; margin: 0; padding: 0; }

body {
  background: var(--bg);
  color: var(--text);
  font-family: var(--sans);
  font-size: 14px;
  line-height: 1.7;
}

/* subtle dot grid */
body::before {
  content: '';
  position: fixed;
  inset: 0;
  background-image: radial-gradient(circle, rgba(120,160,220,0.04) 1px, transparent 1px);
  background-size: 28px 28px;
  pointer-events: none;
  z-index: 0;
}

.page {
  position: relative;
  z-index: 1;
  max-width: 900px;
  margin: 0 auto;
  padding: 3rem 2rem 4rem;
}

/* ── HEADER ── */
.header {
  border-bottom: 1px solid var(--border2);
  padding-bottom: 2rem;
  margin-bottom: 2.5rem;
}

.tag-row {
  display: flex;
  gap: 8px;
  flex-wrap: wrap;
  margin-bottom: 1.2rem;
}

.tag {
  font-family: var(--mono);
  font-size: 10px;
  letter-spacing: 1.2px;
  padding: 3px 10px;
  border-radius: 4px;
  border: 1px solid;
}

.tag-blue  { color: var(--blue);  border-color: rgba(126,184,247,0.3); background: rgba(126,184,247,0.05); }
.tag-teal  { color: var(--teal);  border-color: rgba(86,212,188,0.3);  background: rgba(86,212,188,0.05); }
.tag-yellow{ color: var(--yellow);border-color: rgba(251,191,90,0.3);  background: rgba(251,191,90,0.05); }

h1 {
  font-family: var(--mono);
  font-size: 1.9rem;
  font-weight: 600;
  color: #eef2ff;
  letter-spacing: -0.5px;
  margin-bottom: 0.6rem;
}

h1 span { color: var(--blue); }

.subtitle {
  color: var(--text2);
  font-size: 14px;
  font-weight: 300;
  max-width: 600px;
}

/* ── SECTIONS ── */
section { margin-bottom: 2.5rem; }

h2 {
  font-family: var(--mono);
  font-size: 11px;
  font-weight: 600;
  letter-spacing: 2.5px;
  text-transform: uppercase;
  color: var(--blue);
  margin-bottom: 1.2rem;
  display: flex;
  align-items: center;
  gap: 10px;
}

h2::after {
  content: '';
  flex: 1;
  height: 1px;
  background: var(--border);
}

/* ── PIPELINE ── */
.pipeline {
  display: flex;
  align-items: center;
  gap: 0;
  overflow-x: auto;
  padding: 1.25rem 1.5rem;
  background: var(--card);
  border: 1px solid var(--border);
  border-radius: 12px;
  margin-bottom: 0.75rem;
}

.pipe-step {
  display: flex;
  flex-direction: column;
  align-items: center;
  gap: 5px;
  flex-shrink: 0;
}

.pipe-num {
  font-family: var(--mono);
  font-size: 10px;
  color: var(--text3);
}

.pipe-box {
  background: var(--bg3);
  border: 1px solid var(--border2);
  border-radius: 7px;
  padding: 7px 14px;
  font-family: var(--mono);
  font-size: 11px;
  color: var(--text);
  white-space: nowrap;
}

.pipe-box.highlight {
  border-color: rgba(126,184,247,0.4);
  color: var(--blue);
  background: rgba(126,184,247,0.05);
}

.pipe-arrow {
  font-size: 12px;
  color: var(--text3);
  margin: 0 8px;
  margin-top: 22px;
  flex-shrink: 0;
}

.pipe-note {
  font-size: 10px;
  font-family: var(--mono);
  color: var(--text3);
  text-align: center;
  max-width: 80px;
  line-height: 1.3;
}

/* ── FILE TREE ── */
.file-tree {
  background: var(--card);
  border: 1px solid var(--border);
  border-radius: 12px;
  padding: 1.25rem 1.5rem;
  font-family: var(--mono);
  font-size: 12.5px;
  line-height: 2;
}

.tree-line { display: flex; align-items: center; gap: 0; }
.tree-indent { display: inline-block; width: 20px; color: var(--border2); font-size: 11px; }
.tree-folder { color: var(--yellow); }
.tree-file   { color: var(--text2); }
.tree-key    { color: var(--blue); }
.tree-comment{ color: var(--text3); font-size: 11px; margin-left: 12px; }

/* ── FEATURE TABLE ── */
.feat-table {
  width: 100%;
  border-collapse: collapse;
}

.feat-table th {
  text-align: left;
  font-family: var(--mono);
  font-size: 10px;
  letter-spacing: 1.5px;
  text-transform: uppercase;
  color: var(--text3);
  padding: 8px 14px;
  border-bottom: 1px solid var(--border2);
}

.feat-table td {
  padding: 9px 14px;
  border-bottom: 1px solid var(--border);
  font-size: 13px;
  vertical-align: top;
}

.feat-table tr:last-child td { border-bottom: none; }

.feat-table tbody tr:hover td { background: rgba(126,184,247,0.03); }

.code { font-family: var(--mono); font-size: 11.5px; color: var(--blue); }
.formula { font-family: var(--mono); font-size: 11px; color: var(--teal); }

/* ── MODEL CARDS ── */
.model-grid {
  display: grid;
  grid-template-columns: repeat(auto-fill, minmax(260px, 1fr));
  gap: 12px;
}

.model-card {
  background: var(--card);
  border: 1px solid var(--border);
  border-radius: 10px;
  padding: 1.1rem 1.25rem;
}

.model-card-head {
  display: flex;
  justify-content: space-between;
  align-items: center;
  margin-bottom: 0.7rem;
}

.model-name {
  font-family: var(--mono);
  font-size: 12px;
  font-weight: 600;
  color: var(--text);
}

.model-badge {
  font-family: var(--mono);
  font-size: 9px;
  padding: 2px 7px;
  border-radius: 3px;
  letter-spacing: 0.5px;
}

.badge-supervised { background: rgba(86,212,188,0.1); color: var(--teal); border: 1px solid rgba(86,212,188,0.2); }
.badge-unsupervised{ background: rgba(251,191,90,0.1); color: var(--yellow); border: 1px solid rgba(251,191,90,0.2); }

.model-detail {
  font-size: 12px;
  color: var(--text2);
  line-height: 1.6;
}

.model-detail .kv {
  display: flex;
  justify-content: space-between;
  padding: 2px 0;
  border-bottom: 1px solid var(--border);
}

.model-detail .kv:last-child { border-bottom: none; }
.model-detail .kv .k { color: var(--text3); font-size: 11px; font-family: var(--mono); }
.model-detail .kv .v { font-family: var(--mono); font-size: 11px; color: var(--text); }

/* ── SCORING BLOCK ── */
.score-breakdown {
  background: var(--card);
  border: 1px solid var(--border);
  border-radius: 12px;
  overflow: hidden;
}

.score-row {
  display: grid;
  grid-template-columns: 110px 1fr;
  border-bottom: 1px solid var(--border);
}

.score-row:last-child { border-bottom: none; }

.score-key {
  padding: 11px 16px;
  font-family: var(--mono);
  font-size: 11px;
  color: var(--blue);
  background: rgba(126,184,247,0.04);
  border-right: 1px solid var(--border);
  display: flex;
  align-items: center;
}

.score-val {
  padding: 11px 16px;
  font-family: var(--mono);
  font-size: 11.5px;
  color: var(--text2);
}

.score-val .weight { color: var(--yellow); }
.score-val .op     { color: var(--text3); }

/* ── INSTALL BLOCK ── */
.code-block {
  background: var(--card);
  border: 1px solid var(--border);
  border-radius: 10px;
  padding: 1.1rem 1.5rem;
  font-family: var(--mono);
  font-size: 12.5px;
  line-height: 2;
  overflow-x: auto;
}

.code-block .prompt { color: var(--text3); user-select: none; }
.code-block .cmd    { color: var(--teal); }
.code-block .comment{ color: var(--text3); font-size: 11px; }

/* ── API TABLE ── */
.api-block {
  background: var(--card);
  border: 1px solid var(--border);
  border-radius: 10px;
  overflow: hidden;
}

.api-header {
  display: flex;
  align-items: center;
  gap: 12px;
  padding: 10px 16px;
  border-bottom: 1px solid var(--border);
  background: var(--bg3);
}

.method {
  font-family: var(--mono);
  font-size: 11px;
  font-weight: 600;
  padding: 2px 9px;
  border-radius: 4px;
  background: rgba(86,212,188,0.12);
  color: var(--teal);
  border: 1px solid rgba(86,212,188,0.25);
}

.endpoint {
  font-family: var(--mono);
  font-size: 12px;
  color: var(--text);
}

.api-body { padding: 1rem 1.25rem; }

.field-row {
  display: grid;
  grid-template-columns: 130px 80px 1fr;
  gap: 12px;
  padding: 7px 0;
  border-bottom: 1px solid var(--border);
  font-size: 12px;
}

.field-row:last-child { border-bottom: none; }
.field-name  { font-family: var(--mono); color: var(--blue); }
.field-type  { font-family: var(--mono); color: var(--text3); font-size: 11px; }
.field-desc  { color: var(--text2); }

/* ── FOOTER ── */
footer {
  margin-top: 3rem;
  padding-top: 1.5rem;
  border-top: 1px solid var(--border);
  display: flex;
  justify-content: space-between;
  align-items: center;
  flex-wrap: wrap;
  gap: 0.5rem;
}

footer p {
  font-family: var(--mono);
  font-size: 11px;
  color: var(--text3);
}

.acc-row { display: flex; gap: 8px; flex-wrap: wrap; }
.acc-chip {
  font-family: var(--mono);
  font-size: 11px;
  padding: 3px 9px;
  border-radius: 4px;
  background: rgba(86,212,188,0.07);
  color: var(--teal);
  border: 1px solid rgba(86,212,188,0.15);
}

@media(max-width:600px) {
  .pipeline { flex-direction: column; align-items: flex-start; gap: 6px; }
  .pipe-arrow { transform: rotate(90deg); margin: 0 0 0 30px; }
  .model-grid { grid-template-columns: 1fr; }
  .field-row { grid-template-columns: 1fr 1fr; }
  .field-type { display: none; }
}

</style>

PYTHON 3.13 SCIKIT-LEARN FLASK API SMOTE

Ethereum Fraud Detection

Unsupervised anomaly scoring (Isolation Forest + MF-UFS) used to auto-label Ethereum transactions, followed by supervised classifier training via Logistic Regression, SVM, KNN, Decision Tree, and Random Forest.

Pipeline

01

Feature Engineering

feature_engineering.py

→

02

Isolation Forest

isolation_forest.py

→

03

MF-UFS Scoring

mfufs.py

→

04

Label Creation

create_labels.py

→

05–09

Supervised Models

5 classifiers via main.py

Repository Structure

ml-project-main/

├─main.py— full pipeline runner (steps 1–9)

├─app_backend.py— Flask REST API (port 5000)

├─app.html— browser UI for live predictions

├─eda.py

├─Requirements.txt

├─Data/

│ ├─new dataset/— primary data used by pipeline

│ │ ├─Dataset.csv— raw input

│ │ ├─processed_data.csv— after step 1

│ │ ├─if_scored.csv— after step 2

│ │ ├─final_output.csv— after step 3

│ │ └─labeled_data.csv— after step 4 (model input)

│ └─old dataset/— ethereum.csv + intermediates

├─Models/

│ ├─random_forest.pkl— decision_tree.pkl logistic.pkl

│ ├─svm.pkl knn.pkl— + matching _scaler.pkl files

│ └─decision_tree_plot.png

└─src/

├─preprocessing/ feature_engineering.py

├─models/ isolation_forest.py mfufs.py logistic_model.py svm_model.py knn_model.py decision_tree_model.py random_forest.py

└─evaluation/ create_labels.py

Feature Engineering

Derived from raw columns value, gas, gas_price, receipt_gas_used, block_timestamp, block_number. All five features are z-score normalised before use.

Feature (z-score)	Derived From	Formula
Value_z	value column	value
GasCost_z	gas × gas_price	gas * gas_price
GasEfficiency_z	receipt_gas_used / gas	receipt_gas_used / (gas + ε)
TimeGap_z	block_timestamp diff	diff(block_timestamp).total_seconds()
BlockGap_z	block_number diff	diff(block_number)

Unsupervised Labeling (MF-UFS)

Because no ground-truth labels exist in the dataset, fraud labels are generated automatically by combining three anomaly scores, then thresholding at the top 15%.

IF_Score

Isolation Forest (200 trees, contamination=0.15) — normalised decision function score

StatScore

TempScore

|Value_z − rolling_mean(20)| / rolling_mean — normalised temporal deviation

FinalScore

0.3 × IF_Score + 0.4 × StatScore + 0.3 × TempScore

FraudFlag

1 if FinalScore > 85th percentile, else 0 (≈15% fraud rate)

Models

  <div class="model-card">
    <div class="model-card-head">
      <div class="model-name">Isolation Forest</div>
      <div class="model-badge badge-unsupervised">UNSUPERVISED</div>
    </div>
    <div class="model-detail">
      <div class="kv"><span class="k">n_estimators</span><span class="v">200</span></div>
      <div class="kv"><span class="k">contamination</span><span class="v">0.15</span></div>
      <div class="kv"><span class="k">random_state</span><span class="v">42</span></div>
      <div class="kv"><span class="k">saved</span><span class="v">—&nbsp;(used for scoring only)</span></div>
    </div>
  </div>

  <div class="model-card">
    <div class="model-card-head">
      <div class="model-name">Logistic Regression</div>
      <div class="model-badge badge-supervised">SUPERVISED</div>
    </div>
    <div class="model-detail">
      <div class="kv"><span class="k">max_iter</span><span class="v">1000</span></div>
      <div class="kv"><span class="k">scaler</span><span class="v">StandardScaler</span></div>
      <div class="kv"><span class="k">SMOTE strategy</span><span class="v">auto (balanced)</span></div>
      <div class="kv"><span class="k">saved</span><span class="v">logistic.pkl + scaler</span></div>
    </div>
  </div>

  <div class="model-card">
    <div class="model-card-head">
      <div class="model-name">SVM</div>
      <div class="model-badge badge-supervised">SUPERVISED</div>
    </div>
    <div class="model-detail">
      <div class="kv"><span class="k">kernel</span><span class="v">rbf</span></div>
      <div class="kv"><span class="k">C</span><span class="v">2</span></div>
      <div class="kv"><span class="k">threshold</span><span class="v">0.45 (predict_proba)</span></div>
      <div class="kv"><span class="k">SMOTE strategy</span><span class="v">0.5</span></div>
    </div>
  </div>

  <div class="model-card">
    <div class="model-card-head">
      <div class="model-name">KNN</div>
      <div class="model-badge badge-supervised">SUPERVISED</div>
    </div>
    <div class="model-detail">
      <div class="kv"><span class="k">n_neighbors</span><span class="v">5</span></div>
      <div class="kv"><span class="k">weights</span><span class="v">distance</span></div>
      <div class="kv"><span class="k">scaler</span><span class="v">StandardScaler</span></div>
      <div class="kv"><span class="k">SMOTE strategy</span><span class="v">0.5</span></div>
    </div>
  </div>

  <div class="model-card">
    <div class="model-card-head">
      <div class="model-name">Decision Tree</div>
      <div class="model-badge badge-supervised">SUPERVISED</div>
    </div>
    <div class="model-detail">
      <div class="kv"><span class="k">max_depth</span><span class="v">8</span></div>
      <div class="kv"><span class="k">min_samples_split</span><span class="v">10</span></div>
      <div class="kv"><span class="k">min_samples_leaf</span><span class="v">5</span></div>
      <div class="kv"><span class="k">plot saved</span><span class="v">decision_tree_plot.png</span></div>
    </div>
  </div>

  <div class="model-card">
    <div class="model-card-head">
      <div class="model-name">Random Forest</div>
      <div class="model-badge badge-supervised">SUPERVISED</div>
    </div>
    <div class="model-detail">
      <div class="kv"><span class="k">n_estimators</span><span class="v">300</span></div>
      <div class="kv"><span class="k">max_depth</span><span class="v">12</span></div>
      <div class="kv"><span class="k">min_samples_split</span><span class="v">10</span></div>
      <div class="kv"><span class="k">SMOTE strategy</span><span class="v">0.5</span></div>
    </div>
  </div>

</div>
<p style="margin-top:0.9rem;font-size:12px;color:var(--text3);font-family:var(--mono);">
  All supervised models: 80/20 train-test split · stratified · random_state=42 · SMOTE on training set only.
  SVM and KNN not served via the API (only DT, RF, LR are loaded in app_backend.py).
</p>

Flask API — app_backend.py

POST http://127.0.0.1:5000/predict

REQUEST BODY (JSON)

val_in float ETH received by address

val_out float ETH sent from address

fee float Transaction fee (ETH) — used as GasCost

gas int Gas limit

gas_used int Actual gas used — GasEfficiency = gas_used / gas

time_gap float Seconds since previous transaction

block_gap int Block number difference from previous transaction

Response includes: modelProbs {Decision Tree, Random Forest, Logistic} · FinalScore (mean of three) · IF_Score · StatScore · TempScore · zscores · derivedFeatures

Setup & Run

# 1. Install dependencies

$ pip install -r Requirements.txt

# 2. Run the full pipeline (feature eng → labeling → train all models)

$ python main.py

# 3. Start the Flask API

$ python app_backend.py

# 4. Open app.html in a browser and use the UI

# Make sure the backend is running on localhost:5000

<div style="margin-top:1rem;background:var(--card);border:1px solid var(--border);border-radius:10px;padding:1rem 1.25rem;">
  <p style="font-size:12px;font-family:var(--mono);color:var(--text3);margin-bottom:0.5rem;">REQUIREMENTS.TXT</p>
  <p style="font-family:var(--mono);font-size:12px;color:var(--text2);line-height:2;">
    pandas≥2.0 &nbsp;·&nbsp; numpy≥1.24 &nbsp;·&nbsp; scikit-learn≥1.3 &nbsp;·&nbsp; matplotlib≥3.7
    &nbsp;·&nbsp; seaborn≥0.12 &nbsp;·&nbsp; openpyxl≥3.1 &nbsp;·&nbsp; imbalanced-learn≥0.11
    &nbsp;·&nbsp; streamlit≥1.30 &nbsp;·&nbsp; joblib≥1.3
  </p>
</div>

Ethereum Fraud Detection · MF-UFS Pipeline · CS1138

DT ~93% RF ~95% LR ~88%

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

Repository files navigation

Ethereum Fraud Detection

Pipeline

Repository Structure

Feature Engineering

Unsupervised Labeling (MF-UFS)

Models

Flask API — app_backend.py

Setup & Run

About

Uh oh!

Releases

Packages

Uh oh!

Contributors

Uh oh!

Languages

Name		Name	Last commit message	Last commit date
Latest commit History 62 Commits
Data		Data
Models		Models
Poster		Poster
src		src
Requirements.txt		Requirements.txt
app.html		app.html
app_backend.py		app_backend.py
eda.py		eda.py
main.py		main.py
readme.md		readme.md

Folders and files

Latest commit

History

Repository files navigation

Ethereum Fraud Detection

Pipeline

Repository Structure

Feature Engineering

Unsupervised Labeling (MF-UFS)

Models

Flask API — app_backend.py

Setup & Run

About

Resources

Uh oh!

Stars

Watchers

Forks

Releases

Packages 0

Uh oh!

Contributors

Uh oh!

Languages

Packages