Background:

In digital advertising, not all traffic is of good quality. Industry association uses GIVT (General Invalid Traffic) to refer to a typcial category of invalid traffic which can be defined by a list of data in combination with certain behevior patterns.

The key to define GIVT is a set of IP address black (block) list, whitelist, mobile device ID blacklist, bot and spider user agent list. This challenge focus on building a collaborative GIVT List (IP address and mobile device ID), utilizing blockchain technology and consensus mechanism design to coordinate muiltiple parties' efforts.

The Challenge:

MMA APAC members, over 500 companies in APAC region, intended to build a collaborative database (GIVT List) to define what should be considered invalid traffic and then using this database to filter advertising traffic and results so as to achieve better advertising results and effective cost saving.

The solution is expected to be a system built on enterprise blockchain solution such as Enterprise Ethereum or Hyperledger Fabric. Feature requests:

• Process input from hundreds of companies (nodes of this system)
• Periodically (e.g. hourly, daily) generate censensus results of IP address blacklist、whitelist (and mobile device ID blacklist etc, which follows the same mechanism)
• A voting mechanism to decide what data should go into the data list and what data should be removed from previous version of data list
• The output data list should be stored in a blockchain system and should be decentrialzed and no single node (member) can modify the results without going through consensus mechanism

Assumptions:

• Each member company of this blockchain system will provide a server to run its own node
• Each member company have equal rights in voting
• Each member will submit data to the system from time to time, could be daily or weekly, some may submit data in an ad hoc fashion.
• The data from different companies are of of verious level of data quality and quantity. Not all data will fit the definition of GIVT, therefore a mechanism to count IP address, mobile device ID and use the number of appearance in multiple submition as a proxy of vote. E.g. if IP address 118.118.118.118 appeared in 3 companies submition, this IP address receives 3 votes to get into the database.
• Assuming there will be memership validation process before they can join and setup their own node, the majority node of this network will not initiate attack on the system by submitting low quality data to pollute the database. But as limited by technology capability and data scale or mistake, members will sumbit wrong data accassionally. The system's consensus mechanism should be robust enough to handle certain level of attack.

Specificaitons:

• Data input: list of IPv4 addresses, list of mobile device ID (AAID and IDFA in raw format, lower-case) .
• Each batch of IPv4 addresses or mobile device ID submission is expected to be upto serveral million, depending on each members' GIVT detection capability and business scale. Submission of a couple of thousands or a couple of millions are all normal.
• The solution is preferable to be based on Enterprise Ethereum, the entire system is expected to run with limited centralized managment, which should be limted to the membership approval and other features that must be contralized.
• The out put of this system is a list contains IPv4 addresses that is collaobrative voted as blacklist (same as applies on mobile device ID)

Reference:

A solution, DIF (Distributed Invalid Traffic Filter) to a similar problem is available for reference on Github: https://github.com/tagphi/dif-client