Ad Fraud, Viewability & Brand Safety (IS3150 AY 24/25 Sem 1) PDF

Ad Fraud, Viewability & Brand Safety IS3150 AY 24/25 Sem 1 Professor (Adj) Lai Kok Fung  2 Learning Outcome Understand and learn how to deal with Ad Fraud Ad Viewability Brand Safety and Publisher Ad Quality  Recap:...

Ad Fraud, Viewability & Brand Safety IS3150 AY 24/25 Sem 1 Professor (Adj) Lai Kok Fung  2 Learning Outcome Understand and learn how to deal with Ad Fraud Ad Viewability Brand Safety and Publisher Ad Quality  Recap: Digital Ads Pricing Programmatic Google SEM Affiliate (Display, Social) 3 Recap: Programmatic & RTB 4 Google and Programmatic Other Ad Exchanges DSP SSP DSP RTB SSP SSP Big Advertisers DSP & Agencies Big Publishers DSP Web Interface Java Script Small Advertisers Small Publishers 5 Google  Ad Fraud for Advertisers Programmatic: fully Ad Fraud = automated system with Any deliberate activity that multiple unrelated players & prevents the proper delivery complex ecosystems of ads to the rightful Advertiser intended audience Human or bot/zombie? Human in the right geography or target Pays for Impression Audience group? (CPM) Is the ad viewable? 6  Where do Ad Fraud Occur? Established Ad Exchange Publishers Fraudulent publishers stand to benefit from invalid traffic. Ad exchanges derive about 40% margin from advertising cost. Theoretically, they can be complicit in ad fraud for short- $1 $0.4 $0.6 term gains. Other Ad Exchanges Established players are usually more concerned about long term branding. Choosing alternative ad Non established players may prioritize short-term revenue, exchange for having less stringent fraud Lower cost prevention mechanism. Better performance Alternative Alternative publishers Publishers 7 Not Rightful Intended Audience Dimension Why Fraud What Fraudsters Do Methods Geography First world CPM is higher than Substitute third world traffic for Location Fraud, third world CPM first world traffic Domain Spoofing Class of CPM for high quality publishers Substitute low quality Domain Spoofing, Publishers (e.g. finance) is higher for low publishers for high quality ones User Agent Spoofing quality publishers (e.g. adults) Human or Bot Bot’s traffic can be synthesized in Substitute bots for humans Botnet, with the volume help of malware Viewability More ad view, more revenue Stuff more ads in the same web Pixel stuffing, ad page for real human viewers stacking 8 Location Fraud Paying for First World traffic Getting Third Word traffic instead. 9 Domain Overpaying for lower quality publishers. Also cause brand safety issues. Brand safety = keeping a brand’s reputation safe when advertising Spoofing online, by avoiding placing ads next to inappropriate content. Implications: diminished brand equity, consumer boycott. Fake News 10 Bots: Create Fake Impressions or Clicks Examples: Powerliks – a botnet used for committing both impression and click fraud by generating traffic to ad-supported websites. Avireen – part of a larger malware system called Andromedia 11  (2018) The FBI, Google, and 20 tech industry partners have collaborated to take down a giant cyber-criminal network involved in generating fake ad views and clicks that have been used to defraud ad networks and advertisers for the past International four years and make millions in illicit revenue for the scheme's perpetrators. …. Operations Investigators said that over time, the 3ve operators used different schemes to generate ad views and clicks, relying on a slew of tricks, such as renting other cybercrime botnets, creating their own botnets hosted on commercial data centers, hijacking IP address blocks, using proxies to hide real IP addresses, and even creating their own websites on which they displayed ads, to make sure that bots have ads to load and click on. https://www.scworld.com/news/fbi-swats-down-massive-botnet-fueled-ad-fraud-operation 12  911 S5 Botnet Dismantled and Its Administrator Arrested in Coordinated International Operation https://www.justice.gov/opa/pr/911- s5-botnet-dismantled-and-its- administrator-arrested-coordinated- international-operation Viewability Click Fraud When Payment Model = CPC Examples: Advertisers sabotaging competitors Publishers committing frauds to make illegitimate gains 15 Mobile APP Fraud When Payment Model = CPA Cloud Hosting – mobile devices are hosted on virtual private servers (VPS). In-app impressions are generated by hosting mobile- operating system emulators or devices in the cloud and running apps that display ads, thus generating ad revenue for the app creators. Device Farm – fraudsters manually perform actions such as clicks, installs and other forms of engagement to create the illusion of legitimate activity, draining advertising budgets. 16 The Evolution of Click Farm https://www.youtube.com/ watch?v=-CoEHDHb0lE 17 Advertisers: Dealing with Ad Fraud 18  Stick with the Safe Choices Established Ad Exchange Publishers $1 $0.4 $0.6 Other Ad Exchanges Choosing alternative ad exchange for Lower cost Better performance Alternative publishers Alternative Publishers 19 Ad Fraud Can Also Hurt Publishers 1. CPA Environment Legitimate publishers in a CPA network earn revenue when visitors download an APP. Fraudsters can ”hijack” the attribution process and steal away the credits. 2. Publisher Ad Qualify When undesirable ads sneak through and get shown on publisher sites. 20 Attribution Fraud When another publisher tries to steal credits from another publisher in a CPA environment Attempt to steal credit for app installs, by reporting fake clicks as the last engagement right before the app is first launched by a legitimate user Click flooding – ad networks intentionally report a large number of fraudulent clicks in hopes of taking credit for the last click before an app install happens 21 Ad Quality: for publishers Publishers need to balance protecting their audiences from malicious, disruptive and offensive ads, vs Sacrificing advertising revenue by ruthlessly cutting sources of low- quality demand 22 Example: Ad Sense Control Dealing with Publisher Ad Quality 23 AdSense: Sensitive Categories 24 Full List of Invalid Traffic Types Types of invalid traffic detection, as defined by Pixalate, are reported within the Analytics and Threat Intelligence product solutions. https://www.pixalate.com/knowledgebas e/reported-invalid-traffic-ivt-types … a lot more 25 Ad Fraud: Conclusions Any deliberate activity that prevents the proper delivery of ads to the rightful intended audience Hurts advertisers by siphoning away budget to illegitimate traffic Depresses CPM of legitimate publishers Causes disputes and confusion in the advertiser – exchange – publisher ecosystem. 26 A4: Optional Question 27