|
Quantifying Reputation to Determine 'Do
Most People Want this Message?'
Email authentication can help to eliminate
spoofing and thereby protect the reputation of the righful
owners of IP addresses, domains, email address, and even
message content.
Mailshell offers the first solution specifically
designed to quantify reputation via advanced statistical
models. Our reputation-based analysis creates de facto advanced
'phishing' detection, along with protection from fraud and
other blended threats.
By focusing on reputation, Mailshell is also
able to deploy a language-agnostic statistics-based solution
that catches foreign language spam more accurately than
content-focused solutions.
How We Use 'Reputation' to Catch Spam
We define reputation of an attribute as the
difference between the # of spam versus the # of legit messages.
We call it our SpamRepute Index. For example, the reputation
of an IP address is the difference between the # of spam
messages sent from that IP address versus the # of legit
messages sent from that IP address.
We track the reputation of "hard to
fake" attributes such as:
- Every IP address
- Every domain
- Every sender fingerprint
- Every message fingerprint
To tune the results, we also track the reputation
of related attributes such as:
- Country of origin of IP
- Domain owner
- Name server
- Domain Registrar
Calcluating the Mailshell SpamRepute Index
In practice, when we get a new message, we:
- Extract spam attributes from the message. If they are
spoofed, we throw the message out.
- Compute the SpamRepute Index for each attribute.
- Compute the overall SpamRepute Index for the message
by combining the SpamRepute Index for each attribute.
- Incorporate the overall SpamRepute Index into the overall
spam filtering analysis.
|
