Browse Prior Art Database

Spam / Poison message filtering through phased delivery

IP.com Disclosure Number: IPCOM000236811D
Publication Date: 2014-May-16
Document File: 2 page(s) / 48K

Publishing Venue

The IP.com Prior Art Database

Abstract

A Mechanism is described to avoid the delivery of unwanted messages to wide distribution lists (of users or automated processes) by phasing delivery of the message and allowing early recipients to identify data which should instead be discarded.

This text was extracted from a PDF file.
This is the abbreviated version, containing approximately 47% of the total text.

Page 01 of 2

Spam / Poison message filtering through phased delivery

Messaging systems of all kinds occasionally have a requirement to handle messages which are not 'wanted' by consumers. This applies both to high level user facing systems, such as email (classic 'spam' messages) and application to application middleware messaging systems.

    Many techniques are already well understood which attempt to filter out unwanted messages before they cause harm (ranging from wasting a users time to crashing a critical application). These typically involve pre-parsing the message to some degree, and/or take into account factors such as the source of the message. Some of these techniques can give a definitive answer as to the quality of the message (usually negative, such as 'emails from this domain are explicitly and always blocked', or 'this message will not be correctly parsed by the target application, as it does not include header information X, and we are happy to discard it'. However, often such techniques can only give a probability that the message will be unwanted (e.g. "60% of the time emails which included the words 'cheap', 'medicines', and 'pills' turned out to be spam"). When this is the case, it is usual to 'quarantine' such messages (in a 'Spam folder', or on a 'Dead letter queue') so that human interaction can determine the correct final destination (dispose of the message, or deliver/process as usual). In particular in the case of email messages, some kind of threshold will generally be configured such that when the system is 'X%' certain that message is Spam it can be quarantined, and otherwise it is delivered to the user's Inbox. If this is set anywhere between 0 and 100%, this almost unavoidably means that in some cases there will be false positives (genuinely wanted mail delivered to the 'Spam' folder), and false negatives (Spam

which is not caught, and reaches the Inbox).

In such situations, while the technology significantly improves the end user

experience, there is still a reasonable amount of human intervention required.

    A feature of both of these cases is that they are capable of delivering the same message to multiple consumers. In email terms, this occurs when an email includes multiple addresses in the to/cc/bcc fields, or a distribution list is used. In messaging middleware, distribution lists can again be used, or more importantly perhaps, Publish Subscribe models allow multiple (potentially large numbers) of message consumers to register that they should receive messages on a particular topic.

    Another feature, or potential feature, of both of these domains, is that users/applications receiving unwanted messages can inform the system that this should not have reached them. In many email systems, this is implemented as a 'report Spam' button or similar - messages tagged in this manner will be used to improve future identification of 'Spam' messages. In messaging middleware, an application can typically place a message on a...