Spam Detection Using Sender Trust and Corpus Comparison
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current anti-spam systems in email systems are ineffective in completely preventing spam emails, often resulting in high costs, resource usage, incorrect rejection of legitimate emails, and failure to reject all spam messages.
Innovation Solution
A method that identifies spam by categorizing senders as 'trusted' or 'unknown' based on their relationship with recipients, using a trusted database to filter messages, and comparing message bodies with a corpus of known spam messages to determine spam probability.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If known anti-spam techniques are embedded in email systems, then spam filtering capability is improved, but system cost and resource usage increase
Solution Approach 1:
The system pre-generates multiple candidate reply messages with different reply-times and content variations before the user actually replies. This preliminary generation allows the spam filter to evaluate multiple potential replies simultaneously, improving filtering accuracy without requiring multiple separate filtering operations that would consume additional resources.
Solution Approach 2:
The system creates copy variations of the same reply message with different timing parameters and content slight variations. These copies are used to train and evaluate the spam filter, allowing the system to learn from multiple scenarios without requiring the user to manually create each variation, thus reducing overall system resource consumption while improving reliability.
2Measurement precision
If multiple candidate reply messages are generated with different reply-times, then reply accuracy is improved, but device complexity increases
Solution Approach 1:
The reply generation process is segmented into distinct components: base message content generation, timing parameter assignment, and content variation creation. Each candidate reply is an independent segmented unit with specific timing and content attributes, making the complex process manageable and the system easier to implement and maintain while preserving high reply accuracy.
3Reliability
If spam messages are filtered based on sender address trust, then false positives are reduced, but spam detection accuracy decreases
Solution Approach 1:
The system applies different filtering qualities to different parts of the message evaluation process. Sender address trust is evaluated with one level of scrutiny (reducing false positives), while message body content is evaluated with another level of scrutiny (maintaining detection accuracy). This local differentiation of quality standards allows the system to optimize for both false positive reduction and spam detection accuracy simultaneously.
Data Source
AI summary
The disclosure describes systems and methods for determining spam. A spam detecting system identifies potential spam messages by checking the relationship between a sender of a message and a recipient of the message. The spam detecting system identifies the relationship between the sender and recipient as ‘unknown’ or ‘trusted’. If the message is ‘trusted’, the message is transmitted accordingly to the recipient. If the message is classified as ‘unknown’, then the message is analyzed in view of other identified potential spam messages where a count is tallied against a threshold value to keep track of the probability a message is spam.