Going your own way: A guide to DIY fraud prevention
Posted Sep 7, 2016
With the amount of discussion about fraud both in the market and on the press, it might seem that it’s fairly easy to take on fraud prevention by handling everything yourself, instead of jumping into what is a very competitive and sometimes confusing marketplace. However, rolling your own fraud prevention isn’t easy, even from the get-go: it all depends on your resources, time, and determination. That’s not to say it isn’t doable - far from it, and in this very article we’re going to show you just how easy DIY fraud prevention can be.
The absolute must-haves
There are some aspects of fraud prevention that you should already covered, but let’s take a quick run-through of the basics to see if there’s something that you might have missed:
- Have dedicated (and well-trained!) mobile marketing experts on the team that cover campaign management, Insertion Orders and communication among partners.
- Collaborating closely with advertising partners, and having strong connections with them.
- Make sure you get a solid grounding about your partners’ methods and technology - none of their anti-fraud measures should be a secret, nor too hard to explain. You need to know how your data is being affected by their mechanisms.
- Participate regularly in summits & workshops that will keep you up-to-date with all technical developments, as well as potential new threats. The tech and the threat profile around mobile ad fraud changes all the time.
- Share your experiences with other advertisers. Being a victim of fraud shouldn’t be a stigma - it can even provide you with valuable insights which others can learn from.
You can see that the general theme around this is “training”. Once you have an awareness of the types of fraud that can happen, you might not be able to immediately prevent it or isolate the effects on your data, but you will immediately be able to start recognizing individual risk factors that you can avoid. We ventured to offer this kind of starting point in our whitepaper on mobile publisher fraud.
Ifs and maybes
If you’re often under attack from fraud schemes then you might want to consider one or more of the following solutions:
- Hiring a dedicated anti-fraud professional with extensive knowledge in fraud countermeasures.
- Running checks for symptoms of exploits on your own unified data. Certain unique datapoints might provide additional insights that could have been missed by your partners or tech providers.
- Check your raw data from different sources against each other to pinpoint discrepancies and uncover patterns in traffic or post-install events.
- Create reports on basic quality KPIs per source unique to your campaigns and goals, such as post-install user activity on engagement milestones.
- Try out anti-fraud measures that partners or service providers have successfully employed for you in the past (if you can’t get their service to cover your full advertisement suite.)
Building your own fully automated fraud detection
Pattern recognition, machine learning, data modeling. No matter how sophisticated you build your system, one thing will always be true: your dataset is made up of data points that happened in the past. That means, by identifying fraud, you can only ever react to it after the fact.
But remember, most technical and compliance schemes can be prevented by your tracking and analytics provider (and digital advertising security services) at a fraction of the cost of building and maintaining your own detection system. As such, keep an eye on the market, as you could save on time and money in the process.
Starting your own blacklist
You may be considering building your own blacklist of publishers, exchanges or IP addresses, but there are some pitfalls that could cause you harm with this method. Lets look at the potential problems of blacklisting, just to give you something to think about before taking the plunge.
Blacklisting IPs sounds easy, but in reality getting it right is like threading a needle blindfolded. For starters, IP addresses of clicks and impressions do not always match the addresses of installs. This one of the reasons why fingerprinting that’s reliant on an IP address is just a fallback method - the bulk of tracking is handled via device identifiers and app store referrers. So, the first problem would be to find the correct IP address that you want to blacklist.
The second problem is to decide whether to blacklist an IP address which is incriminated in a fraud case, as there’s always a risk of false positives. Also, without intricate knowledge of the the metadata around an IP address (such as country, ISP and organization) you might blacklist IPs from shared pools for mobile devices used by carriers. Some carriers use so-called “NATting” services that reduce the amount of IP addresses a carrier has to utilize to route mobile traffic through the internet. Those addresses might look like they belong to a data center, but in fact are actually leased out to a mobile carrier instead.
Buying IP blacklists to report on that aren’t suited to your purpose will also net false results. For instance, databases that contain IP addresses incriminated in ecommerce will naturally include residential IP addresses that belong to broadband or mobile carriers, and those lists have no application for flagging fraud in mobile app installs, since both should exclusively originate from residential and carrier IP addresses.
Publisher, network or exchange blacklisting
It might seem like a good idea to create a blacklist of people you don’t want to work with again, should your relationship be harmed by fraud. However, this might not actually reduce your exposure to fraud. This is because the people generating their income out of exploiting campaigns don’t work in agencies, but instead find ways of manipulating the system. These criminals make sure that they have no direct partnerships to networks or advertisers who could be at risk of prosecution. Blacklisting partners would reduce your pool of partners, more than anything else. Exploiters have plenty of partnerships with low and mid-level ad-networks, and can spread out their operation as far as possible to protect their income. These sources might pop up on the next network you work with as soon as you blacklisted the first.
We’re not saying that you shouldn’t hold your partners accountable for their actions, but overzealous blacklisting could hurt your reach, potential volumes and CPx prices without adding any fraud security. Blacklisters, be warned.
Still thinking about rolling your own fraud prevention?
Though this can take some time, if it works well the rewards could be very great. If you want to find out more about fraud prevention, follow the link to our Mobile Publisher Fraud whitepaper here. Alternatively, go back to basics with our mobile fraud guide, which covers the various sources of ad fraud, how they affect your campaigns, and what you can do to fight them.