init commit

2026-07-29 16:50:22 +03:00
commit 0d10a278bf
24 changed files with 3564 additions and 0 deletions
--- a/lab5/generate_http_data.py
+++ b/lab5/generate_http_data.py
@@ -0,0 +1,56 @@
+import csv
+import random
+
+random.seed(42)
+
+MALICIOUS_TEMPLATES = [
+    "<script>alert('xss')</script>",
+    "<img src=x onerror=alert(1)>",
+    "<svg onload=alert(1)>",
+    "javascript:alert(document.cookie)",
+    "<iframe src='javascript:alert(1)'></iframe>",
+    "'; system('rm -rf /')",
+    "$(cat /etc/passwd)",
+    "`whoami`",
+    "<body onload=alert('xss')>",
+    "../../../../etc/passwd",
+    "<a href=\"javascript:void(0)\" onclick=\"alert(1)\">click</a>",
+    "%3Cscript%3Ealert(1)%3C/script%3E",
+    "; ping -c 10 127.0.0.1;",
+    "<input onfocus=alert(1) autofocus>",
+]
+
+BENIGN_TEMPLATES = [
+    "search term: red shoes size {n}",
+    "username: alex_{n}",
+    "comment: I really liked this product, arrived in {n} days",
+    "city: Springfield",
+    "message: see you at {n}pm tomorrow",
+    "review title: Great value for {n} dollars",
+    "address line 2: Apt {n}B",
+    "notes: please call before {n}pm",
+    "tag: outdoor, hiking, {n}-season",
+    "bio: software engineer, {n} years experience",
+]
+
+
+def fill(template):
+    return template.format(n=random.randint(2, 30))
+
+
+def generate(n_per_class=250):
+    rows = []
+    for _ in range(n_per_class):
+        rows.append((random.choice(MALICIOUS_TEMPLATES), 1))
+        rows.append((fill(random.choice(BENIGN_TEMPLATES)), 0))
+    random.shuffle(rows)
+    return rows
+
+
+if __name__ == "__main__":
+    rows = generate(n_per_class=250)
+    with open("data.csv", "w", newline="", encoding="utf-8") as f:
+        writer = csv.writer(f)
+        writer.writerow(["text", "label"])
+        writer.writerows(rows)
+    print(f"Wrote {len(rows)} rows to data.csv")
--- a/lab5/generate_sql_data.py
+++ b/lab5/generate_sql_data.py
@@ -0,0 +1,60 @@
+import csv
+import random
+
+random.seed(42)
+
+SQLI_TEMPLATES = [
+    "' OR '1'='1",
+    "' OR 1=1 --",
+    "admin'--",
+    "' UNION SELECT username, password FROM users--",
+    "1; DROP TABLE users;--",
+    "' AND SLEEP(5)--",
+    "\" OR \"\"=\"",
+    "1' AND '1'='1' UNION SELECT NULL,NULL--",
+    "'; EXEC xp_cmdshell('dir')--",
+    "' OR 'a'='a",
+    "1 OR 1=1",
+    "' OR EXISTS(SELECT * FROM users)--",
+    "%27%20OR%20%271%27%3D%271",
+    "'/**/OR/**/1=1--",
+]
+
+BENIGN_TEMPLATES = [
+    "john.doe@example.com",
+    "Jane O'Brien",
+    "New York, NY",
+    "Sunset Boulevard Apt 4B",
+    "search term: winter jackets",
+    "user123",
+    "2024-05-19",
+    "5 stars, would buy again",
+    "quantity: 3",
+    "order #{n}",
+    "product review for {n}",
+    "phone: 555-01{n}",
+    "comment: great service, thanks!",
+    "coupon code SAVE{n}",
+]
+
+
+def fill(template):
+    return template.format(n=random.randint(10, 99))
+
+
+def generate(n_per_class=250):
+    rows = []
+    for _ in range(n_per_class):
+        rows.append((random.choice(SQLI_TEMPLATES), 1))
+        rows.append((fill(random.choice(BENIGN_TEMPLATES)), 0))
+    random.shuffle(rows)
+    return rows
+
+
+if __name__ == "__main__":
+    rows = generate(n_per_class=250)
+    with open("data.csv", "w", newline="", encoding="utf-8") as f:
+        writer = csv.writer(f)
+        writer.writerow(["text", "label"])
+        writer.writerows(rows)
+    print(f"Wrote {len(rows)} rows to data.csv")
--- a/lab5/test_sqli.py
+++ b/lab5/test_sqli.py
@@ -0,0 +1,15 @@
+import joblib
+
+vectorizer = joblib.load("sqli_vectorizer.joblib")
+clf = joblib.load("sqli_classifier.joblib")
+
+print("Type text to classify. Ctrl+C or empty line to quit.\n")
+while True:
+    text = input("> ").strip()
+    if not text:
+        break
+    X = vectorizer.transform([text])
+    pred = clf.predict(X)[0]
+    verdict = "MALICIOUS" if pred == 1 else "benign"
+    print(f"[{verdict}] {text}\n")
+
--- a/lab5/test_xss.py
+++ b/lab5/test_xss.py
@@ -0,0 +1,15 @@
+import joblib
+
+vectorizer = joblib.load("xss_vectorizer.joblib")
+clf = joblib.load("xss_classifier.joblib")
+
+print("Type text to classify. Ctrl+C or empty line to quit.\n")
+while True:
+    text = input("> ").strip()
+    if not text:
+        break
+    X = vectorizer.transform([text])
+    pred = clf.predict(X)[0]
+    verdict = "MALICIOUS" if pred == 1 else "benign"
+    print(f"[{verdict}] {text}\n")
+
--- a/lab5/train_http_model.py
+++ b/lab5/train_http_model.py
@@ -0,0 +1,77 @@
+import joblib
+import csv
+import sys
+from pathlib import Path
+
+import numpy as np
+from scipy.sparse import hstack
+from sklearn.feature_extraction.text import TfidfVectorizer
+from sklearn.metrics import accuracy_score, classification_report, confusion_matrix
+from sklearn.model_selection import train_test_split
+from sklearn.svm import LinearSVC
+
+sys.path.append(str(Path(__file__).resolve().parent.parent))
+from common.features import handcrafted_feature_matrix  # noqa: E402
+
+XSS_KEYWORDS = [
+    "<script", "onerror", "onload", "javascript:", "onclick", "onfocus",
+    "<iframe", "<svg", "alert(", "system(", "etc/passwd", "cmd", "eval(",
+]
+
+
+def load_data(path="data.csv"):
+    texts, labels = [], []
+    with open(path, newline="", encoding="utf-8") as f:
+        for row in csv.DictReader(f):
+            texts.append(row["text"])
+            labels.append(int(row["label"]))
+    return texts, np.array(labels)
+
+
+def main():
+    texts, labels = load_data()
+    X_train_txt, X_test_txt, y_train, y_test = train_test_split(
+        texts, labels, test_size=0.25, random_state=42, stratify=labels
+    )
+
+    vectorizer = TfidfVectorizer(analyzer="char_wb", ngram_range=(2, 4), max_features=1500)
+    Xtr_tfidf = vectorizer.fit_transform(X_train_txt)
+    Xte_tfidf = vectorizer.transform(X_test_txt)
+
+    Xtr_hand = handcrafted_feature_matrix(X_train_txt, XSS_KEYWORDS)
+    Xte_hand = handcrafted_feature_matrix(X_test_txt, XSS_KEYWORDS)
+
+    X_train = hstack([Xtr_tfidf, Xtr_hand])
+    X_test = hstack([Xte_tfidf, Xte_hand])
+
+    clf = LinearSVC(random_state=42, max_iter=5000)
+    clf.fit(X_train, y_train)
+    y_pred = clf.predict(X_test)
+
+    print("HTTP / XSS Injection Detection")
+    print(f"Accuracy: {accuracy_score(y_test, y_pred):.3f}")
+    print(classification_report(y_test, y_pred, target_names=["benign", "malicious"]))
+    print("Confusion matrix [[TN FP][FN TP]]:")
+    print(confusion_matrix(y_test, y_pred))
+
+    print("\nTest Samples")
+    samples = [
+        "<script>alert(document.cookie)</script>",
+        "search term: comfortable running shoes",
+        "<img src=x onerror=alert(1)>",
+        "notes: please call before 5pm",
+    ]
+    Xd_tfidf = vectorizer.transform(samples)
+    Xd_hand = handcrafted_feature_matrix(samples, XSS_KEYWORDS)
+    Xd = hstack([Xd_tfidf, Xd_hand])
+    preds = clf.predict(Xd)
+    for text, label in zip(samples, preds):
+        verdict = "MALICIOUS" if label == 1 else "benign"
+        print(f"[{verdict}] {text}")
+
+
+if __name__ == "__main__":
+    joblib.dump(vectorizer, "xss_vectorizer.joblib")
+    joblib.dump(clf, "xss_classifier.joblib")
+    print("\nSaved model artifacts. Run: python test_model.py")
+    main()
--- a/lab5/train_sqli_model.py
+++ b/lab5/train_sqli_model.py
@@ -0,0 +1,77 @@
+import joblib
+import csv
+import sys
+from pathlib import Path
+
+import numpy as np
+from scipy.sparse import hstack
+from sklearn.ensemble import RandomForestClassifier
+from sklearn.feature_extraction.text import TfidfVectorizer
+from sklearn.metrics import accuracy_score, classification_report, confusion_matrix
+from sklearn.model_selection import train_test_split
+
+sys.path.append(str(Path(__file__).resolve().parent.parent))
+from common.features import handcrafted_feature_matrix  # noqa: E402
+
+SQL_KEYWORDS = [
+    "select", "union", "drop", "insert", "update", "delete", "or 1=1",
+    "sleep(", "xp_cmdshell", "--", "exec ", "'or'", "1=1",
+]
+
+
+def load_data(path="data.csv"):
+    texts, labels = [], []
+    with open(path, newline="", encoding="utf-8") as f:
+        for row in csv.DictReader(f):
+            texts.append(row["text"])
+            labels.append(int(row["label"]))
+    return texts, np.array(labels)
+
+
+def main():
+    texts, labels = load_data()
+    X_train_txt, X_test_txt, y_train, y_test = train_test_split(
+        texts, labels, test_size=0.25, random_state=42, stratify=labels
+    )
+
+    vectorizer = TfidfVectorizer(analyzer="char_wb", ngram_range=(2, 4), max_features=1500)
+    Xtr_tfidf = vectorizer.fit_transform(X_train_txt)
+    Xte_tfidf = vectorizer.transform(X_test_txt)
+
+    Xtr_hand = handcrafted_feature_matrix(X_train_txt, SQL_KEYWORDS)
+    Xte_hand = handcrafted_feature_matrix(X_test_txt, SQL_KEYWORDS)
+
+    X_train = hstack([Xtr_tfidf, Xtr_hand])
+    X_test = hstack([Xte_tfidf, Xte_hand])
+
+    clf = RandomForestClassifier(n_estimators=200, random_state=42)
+    clf.fit(X_train, y_train)
+    y_pred = clf.predict(X_test)
+
+    print("SQL Injection Detection")
+    print(f"Accuracy: {accuracy_score(y_test, y_pred):.3f}")
+    print(classification_report(y_test, y_pred, target_names=["benign", "malicious"]))
+    print("Confusion matrix [[TN FP][FN TP]]:")
+    print(confusion_matrix(y_test, y_pred))
+
+    print("\nTest Samples")
+    samples = [
+        "' OR '1'='1' --",
+        "jane.doe@example.com",
+        "1; DROP TABLE orders;--",
+        "search: blue running shoes size 10",
+    ]
+    Xd_tfidf = vectorizer.transform(samples)
+    Xd_hand = handcrafted_feature_matrix(samples, SQL_KEYWORDS)
+    Xd = hstack([Xd_tfidf, Xd_hand])
+    preds = clf.predict(Xd)
+    for text, label in zip(samples, preds):
+        verdict = "MALICIOUS" if label == 1 else "benign"
+        print(f"[{verdict}] {text}")
+
+
+if __name__ == "__main__":
+    joblib.dump(vectorizer, "sqli_vectorizer.joblib")
+    joblib.dump(clf, "sqli_classifier.joblib")
+    print("\nSaved model artifacts. Run: python test_model.py")
+    main()