Spam Classification via Network Flow Metadata

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

Conventional spam and malicious email detection methods, such as content-based filters and DNS blackhole lists, are ineffective against evolving spam tactics and can mistakenly block legitimate emails, while detecting spam from network metadata is challenging due to the lack of content information.

Innovation Solution

A machine learning-based spam classification system that utilizes IPFIX data from cloud service networks to identify hidden patterns in network metadata features like external IPs, ports, and TCP flags, trained with labels from email service providers to dynamically adapt to changing spam patterns, allowing for spam detection without accessing email content.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Reliability

If content-based filters are used to detect spam, then spam detection capability is improved, but legitimate emails are mistakenly blocked

Engineering Contradiction:
Improvespam detection capabilityVSAvoidfalse positive blocking
Core Design Contradiction:
ReliabilityVSObject-affected harmful factors

Solution Approach 1:

The patent introduces network metadata as an intermediary feature set that bridges the gap between content-based detection and network-level filtering. By extracting features like IP addresses, ports, and TCP flags from network metadata, the system creates a mediating layer that provides spam detection signals without directly analyzing email content, thus reducing false positives while maintaining detection capability

Inventive Principle:
Principle #24Intermediary (Mediator)

Solution Approach 2:

The patent replaces the mechanical content-based filtering system with a network metadata-based detection system. Instead of analyzing email content directly (mechanical approach), the system substitutes this with analyzing network flow data and metadata patterns, which indirectly reveal spam characteristics without the false positive problems of content-based filters

Inventive Principle:
Principle #28Mechanics substitution (Replace mechanical system)

2Reliability

If DNS blackhole lists are used to block spam, then spam blocking is improved, but legitimate emails from shared servers are blocked

Engineering Contradiction:
Improvespam blocking effectivenessVSAvoidlegitimate email delivery
Core Design Contradiction:
ReliabilityVSProductivity

Solution Approach 1:

The patent applies local quality by transitioning from global IP-level blocking to more granular, context-specific detection. Instead of blocking all emails from an IP address (global approach), the system analyzes specific network metadata features and communication patterns for each email, allowing legitimate emails from shared servers to pass through while blocking only the actual spam

Inventive Principle:
Principle #3Local quality

Solution Approach 2:

The patent changes the detection parameters from static IP address lists to dynamic network metadata features including IP addresses, ports, TCP flags, and communication patterns. This parameter change enables the system to distinguish between legitimate and spam traffic more accurately, maintaining productivity while improving spam blocking

Inventive Principle:
Principle #35Parameter changes

3Measurement precision

If conventional content-based filters are used, then exact match detection is improved, but spam with modified content evades detection

Engineering Contradiction:
Improveexact match detectionVSAvoidresistance to spam modification
Core Design Contradiction:
Measurement precisionVSAdaptability or versatility

Solution Approach 1:

The patent applies preliminary action by analyzing network metadata and communication patterns before the email content is fully processed or even delivered. By examining IP addresses, ports, and TCP flags in advance, the system establishes a baseline of suspicious behavior that remains detectable even when spam content is modified, providing early warning and detection capability

Inventive Principle:
Principle #10Preliminary action

Solution Approach 2:

The patent introduces dynamics by using machine learning models that continuously adapt to changing spam patterns. The system dynamically adjusts its detection criteria based on learned patterns from network metadata, allowing it to maintain measurement precision while becoming increasingly adaptable to modified spam content over time

Inventive Principle:
Principle #15Dynamics

Data Source

PatentEP3469770B1Spam classification system based on network flow data
Publication Date: 2021.05.19 MICROSOFT TECHNOLOGY LICENSING LLC
  • EP3469770B1 patent drawingFigure 1
  • EP3469770B1 patent drawingFigure 2
  • EP3469770B1 patent drawingFigure 3

AI summary

In an example embodiment, a computer-implemented method comprises obtaining labels from messages associated with an email service provider, wherein the labels indicate for each message IP how many spam and non-spam messages have been received; obtaining network data features from a cloud service provider; providing the labels and network data features to a machine learning application; generating a prediction model representing an algorithm for determining whether a particular set of network data features are spam or not; applying the prediction model to network data features for an unlabeled message; and generating an output of the prediction model indicating a likelihood that the unlabeled message is spam.