sameersyed/Defence_FrameWork
0
1�
2 �
i� � �\ � S SK r S SKrS SKrS SKJr S SKrS SKJ r S SK3J r " S S5 r g)� N)�Image)� MobileNetV2)�preprocess_inputc �& � \ rS rSrS rS rS rSrg)�ImageSpamDetector� c �< � [ SSSS9U l SSSSS.U l g ) N�imagenetF�avg)�weights� include_top�pooling� � � )�
bright_colors�
high_contrast�4small_size�
ad_dimensions)r �model�
spam_patterns)�selfs �+D:\Security_app\backend\spam_detector_ml.py�__init__�ImageSpamDetector.__init__5 s+ � � ���PU�V��6� � � � � 7��� c
� � SU; a UR S5 S n[ R " U5 n[ R " [8 R " U5 5 R S5 nUR u pE[ R " UR S5 5 n[ R " USS9n[ U5 nU R R USS9S nSn/ n XE4S; a US -
nU R! S95 XE- n10U11S :� d U12S : a US
-
nU R! S5 [ R " U5 n [ R" " U 5 n U S:� a US-
nU R! S5 [ R$ " U 5 n
U
S:� a US-
nU R! S5 [ R& " U5 nUS: a US-
nU R! S5 US:� a SnSnOUS :� a SnSnOS nS n[) US 5 UUU UUUR* [- [/ U 5 S 5 [- [/ U
5 S 5 S .S!.$ ! [0 a nS"S#[3 U5 30s S nA$ S nAff = f)$N�,� �RGB)�� r! r )�axis)�verbose))i� �Z )i, �� )i� �< )� iX r zCommon ad size detected� g333333�?r zUnusual aspect ratio�� � z"Very bright colors (common in ads)�F zHigh contrast detectedg{�G�z�?�13 zLow visual complexity�2 zHIGH SPAM RISK�redzMEDIUM SPAM RISK�orangez
LOW SPAM RISK�green�d � )�width�height�format�14brightness�contrast)�15spam_score�result�color�16indicators�17image_info�errorzFailed to analyze: )�split�base64� b64decoder �open�io�BytesIO�convert�size�np�array�resize� expand_dimsr r �predict�append�mean�std�var�minr5 �round�float� Exception�str)r �18image_data� image_bytes�imager3 r4 � img_array�featuresr8 r; � aspect_ratio�
img_array_rgbr6 r7 �feature_variancer9 r: �es r �
analyze_image� ImageSpamDetector.analyze_image s] � �K =��j� � '� -� -�c� 2�1� 5�19� �*�*�:�6�K��J�J�r�z�z�+� 6�7�?�?��F�E� !�J�J�M�E� ���� � �j�!9�:�I����y�q�9�I�(��3�I��z�z�)�)�)�Q�)�?��B�H� �J��J� ��"P�P��b� �20��!�!�";�<� !�>�L��a� �<�#�#5��b� �21��!�!�"8�9� �H�H�U�O�M���� �/�J��C� ��b� �22��!�!�"F�G� �v�v�m�,�H��"�}��b� �23��!�!�":�;� "�v�v�h� /� � �$�&��b� �24��!�!�"9�:� �R� �)�� ���r�!�+�� ��(�� �� "�*�c� 2� � � (� "� $� #�l�l�"'��j�(9�1�"=� %�e�H�o�q� 9� � �
�� � =�� 2�3�q�6�(� ;�<� <�� =�s �H-H0 �025I�:I� I�Ic �6 � SU; a SUS 3$ SUS SUS SUS S S26US S S US S
SUS S SUS S S3nUS ( a US H nUSU 3-
nM
OUS-
nUS-
nUS S:� a US-
nU$ US S:� a US-
nU$ US-
nU$ ) Nr= zError: zY27## Image Spam Analysis Report (ML-Enhanced)28 29### Spam Risk Assessment30 31**Risk Level:** `r9 z` (r8 u7 /100)32 33---34 35### Image Information36 37• **Dimensions:** r< r3 z x r4 u pixels38• **Format:** r5 u 39• **Brightness:** r6 u /25540• **Contrast:** r7 z41 42---43 44### Spam Indicators45r; u 46• u 47✓ No spam indicators detectedz48 49---50 51### Recommendation52 53r- zF**High Risk:** Strong spam characteristics detected using ML analysis.r zC**Medium Risk:** Some spam characteristics detected. Verify source.z<**Low Risk:** Image appears legitimate based on ML analysis.� )r �analysis�report� indicators r �generate_report�!ImageSpamDetector.generate_reportd sE � � �h� � �X�g� .� /�0� 0��54 �8�$�%�S��,�)?�(@� A� �l�+�G�4�5�S��,�9O�PX�9Y�8Z� [��,�'��1�2� 3� �l�+�L�9�:� ;��L�)�*�5�6� 7� ��* �L� !� %�l� 3� ��F�9�+�.�.�� 4�
�9� 9�F��5�5�� �L� !�R� '� �^� ^�F� �
� �l�
#�r�
)� �[� [�F� �
�
�T� T�F��
r )r r N)�__name__�55__module__� __qualname__�__firstlineno__r r] rd �__static_attributes__r` r r r r s � �5657�L=�\(r r )
r? rB �numpyrF �PILr �58tensorflow�tf� tensorflow.keras.applicationsr �*tensorflow.keras.applications.mobilenet_v2r r r` r r �<module>rq s& ��
� � � � � 5� G�C� Cr