Create anti-malware application that uses ML, wikipedia bot, WikiLLM and RAG boilerplate

Python/Anti-Malware_application/README.md

@@ -0,0 +1,26 @@
+# A ML-based anti-malware program written in Python #
+
+<p> I first created this project for a competition. I had to learn a lot of new stuff and take inspiration(and a bit of code) from others who had tried similar projects before. This project uses customtkinter and scikit-learn to handle the GUI and ML operations respectively. It utilizes a decision tree classifier to classify files as malware or benign. It can classify a single file or all the files in a folder. It currently only works with executable files(but can be modified to work with other file types). Using the PEfile library, this program extracts information from the executable file(s) to classify them. 
+
+
+Note: It is not perfect and is prone to a lot of false-positives, but I hardly encountered any false-negatives. I think this is due to the overfitting of the model.
+
+### Installation and execution
+
+1) Install the necessary libraries
+```
+pip3 install customtkinter
+pip3 install tkinter
+pip3 install pandas
+pip3 install sklearn
+pip3 install pefile
+pip3 install numpy
+
+```
+
+2) Execute the program
+```
+python3 anti-malware.py
+```
+
+Happy learning!

Python/Anti-Malware_application/anti-malware.py

@@ -0,0 +1,241 @@
+#################################################################################
+###   Author:  Pyerie                         									#
+###   Application: A not-so-accurate ML based anti-malware solution 			# 
+#################################################################################
+
+print("[+] Loading.... ")
+import customtkinter
+from tkinter.filedialog import *
+from tkinter import *
+import pefile 
+import numpy as np
+import pandas as pd 
+from sklearn.tree import DecisionTreeClassifier
+from sklearn.model_selection import train_test_split
+from sklearn import metrics
+import os
+
+
+
+dataset = pd.read_csv('database3.csv')
+X = dataset.drop(['legitimate'],axis=1).values
+
+y = dataset['legitimate'].values
+
+
+
+X_train, X_test, y_train, y_test = train_test_split(X, y, test_size=0.3, random_state=1)
+clf = DecisionTreeClassifier()
+
+
+y_test = y_test.reshape(-1,1)
+for i in range(0, 10):
+	clf = clf.fit(X_train,y_train)
+res1 = clf.predict(X_test)
+accuracy = metrics.accuracy_score(y_test, res1)
+accuracy = str(accuracy)[2:4] + "%"
+print("Accuracy: "+accuracy)
+
+
+customtkinter.set_appearance_mode("dark")
+customtkinter.set_default_color_theme("dark-blue")
+
+
+window = Tk()
+screen_width = window.winfo_screenwidth()
+screen_height = window.winfo_screenheight()
+window.geometry(str(screen_width)+"x"+str(screen_height))
+window.title("eSuraksha")
+window['bg'] = "#121212"
+def extract_features(file):
+	features = []
+
+
+
+	try:
+
+		pe_obj = pefile.PE(file, fast_load=True)
+	except pefile.PEFormatError as error:
+		print("Not PE file!")
+
+	features.append(pe_obj.OPTIONAL_HEADER.DATA_DIRECTORY[6].Size)
+	features.append(pe_obj.OPTIONAL_HEADER.DATA_DIRECTORY[6].VirtualAddress)
+	features.append(pe_obj.OPTIONAL_HEADER.MajorImageVersion)
+	features.append(pe_obj.OPTIONAL_HEADER.MajorOperatingSystemVersion)
+	features.append(pe_obj.OPTIONAL_HEADER.DATA_DIRECTORY[0].VirtualAddress)
+	features.append(pe_obj.OPTIONAL_HEADER.DATA_DIRECTORY[0].Size)
+	try:
+	    features.append(pe_obj.OPTIONAL_HEADER.DATA_DIRECTORY[12].VirtualAddress)
+	except:
+	  	features.append(0)
+	features.append(pe_obj.OPTIONAL_HEADER.DATA_DIRECTORY[2].Size)
+	features.append(pe_obj.OPTIONAL_HEADER.MajorLinkerVersion)
+	features.append(pe_obj.FILE_HEADER.NumberOfSections)
+	features.append(pe_obj.OPTIONAL_HEADER.SizeOfStackReserve)
+	features.append(pe_obj.OPTIONAL_HEADER.DllCharacteristics)
+	features.append(pe_obj.OPTIONAL_HEADER.AddressOfEntryPoint)
+	features.append(pe_obj.OPTIONAL_HEADER.ImageBase)
+
+
+
+
+
+
+	return features
+
+toplevel_created = False
+
+toplevel2_created = False
+
+def single_file():
+
+	global toplevel_created
+	global toplevel2_created
+	global single_file_top
+	if toplevel_created == "True":
+		single_file_top.destroy()
+		toplevel_created = "False"
+	elif toplevel_created == "False":
+		pass
+
+	if toplevel2_created == "True":
+		many_files.destroy()
+		toplevel2_created = "False"
+	elif toplevel2_created == "False":
+		pass
+
+	single_file_top = Toplevel(window)
+	single_file_top.geometry("350x200")
+	customtkinter.set_appearance_mode("dark")
+	customtkinter.set_default_color_theme("dark-blue")
+	single_file_top['bg'] = "#121212"
+	single_file_top.title("Scan a single file")
+	toplevel_created = "True"
+	result = customtkinter.CTkLabel(single_file_top, text="Loading...")
+	result.pack()
+
+	file_path = askopenfilename()
+	try:
+		features_extracted = extract_features(str(file_path))
+		not_pe = False
+	except UnboundLocalError as e:
+		not_pe = True
+		result.after(0, result.destroy)
+		benign_l = customtkinter.CTkLabel(single_file_top,  text="Not PE file!")
+		benign_l.pack()
+		toplevel2_created = False
+
+	if not_pe != True:
+		data_of_sample = np.array(features_extracted)
+		data_of_sample = data_of_sample.reshape(1,-1)
+
+
+		prediction = clf.predict(data_of_sample)
+
+
+		if prediction == 1:
+			result.after(0, result.destroy)
+
+			malware_l = customtkinter.CTkLabel(single_file_top, fg_color="red", text="ML model detected malware!")
+			malware_l.pack()
+
+
+		elif prediction == 0:
+			result.after(0, result.destroy)
+			benign_l = customtkinter.CTkLabel(single_file_top, fg_color="green", text="No malware detected!")
+			benign_l.pack()
+
+
+def scan_many():
+
+
+	global toplevel2_created
+	global toplevel_created
+	global many_files
+
+	if toplevel2_created == "True":
+		many_files.destroy()
+		toplevel2_created = "False"
+	elif toplevel2_created == "False":
+		pass
+
+	if toplevel_created == "True":
+		single_file_top.destroy()
+		toplevel_created = "False"
+	elif toplevel_created == "False":
+		pass
+
+	many_files = Toplevel(window)
+	many_files.geometry("350x200")
+	customtkinter.set_appearance_mode("dark")
+	customtkinter.set_default_color_theme("dark-blue")
+	many_files['bg'] = "#121212"
+	many_files.title("Scan a directory")
+	toplevel2_created = "True"
+	result2 = customtkinter.CTkLabel(many_files, text="Loading...")
+	result2.pack()
+	malware_many = []
+	directory = askdirectory()
+	global extracted
+	for root, directory, files in os.walk(str(directory)):
+			for name_of_file in files:
+				path = os.path.join(str(root),str(name_of_file))
+
+				formats_of_pe = [".acm" , ".ax" , ".cpl" , ".dll" , ".drv" , ".efi" , ".exe" , ".mui" , ".ocx" , ".scr" , ".sys" , ".tsp", ".bin"]
+				for format_i in formats_of_pe:
+					if name_of_file.endswith(format_i) == True:
+
+						extracted = 1
+						try:
+
+							features_of_many = extract_features(str(path))
+						except UnboundLocalError as e:
+							pass
+						break
+
+					else:
+						extracted = 0
+
+
+
+				if extracted == 1:
+					data_for_many = np.array(features_of_many)
+					data_for_many = data_for_many.reshape(1,-1)
+
+					prediction_for_many = clf.predict(data_for_many)
+
+
+					if prediction_for_many == 1:
+						malware_many.append(str(path))
+
+
+	if len(malware_many) != 0:
+		result2.after(0, result2.destroy)
+		malware_label2 = customtkinter.CTkLabel(many_files,text="Malware found: ")
+		malware_label2.pack()
+		malware_text_box = customtkinter.CTkTextbox(many_files)
+		for_text_box = ''
+
+		for name_of_malware in malware_many:
+			for_text_box += "".join([name_of_malware, '\n------------------------------------------']) 
+
+
+
+		malware_text_box.insert('0.0',for_text_box)
+		malware_text_box.configure(state="disabled")
+		malware_text_box.pack() 
+
+
+
+
+	elif len(malware_many) == 0:
+		result2.after(0, result2.destroy)
+		benign_label = customtkinter.CTkLabel(many_files,text="No malware found!")
+		benign_label.pack()
+
+button1 = customtkinter.CTkButton(master=window, command=single_file,text="Scan a single file")
+button1.pack()
+button2 = customtkinter.CTkButton(master=window, command=scan_many, text="Scan a folder")
+button2.pack()
+
+window.mainloop()