Détection du trafic invalide
Le CESP souligne que cette grille est basée sur la confiance accordée par le CESP à l’égard des informations fournies par les mesureurs.
Les réponses indiquées dans la grille ont été communiquées en anglais, par les équipes techniques de chaque mesureur.
Pour naviguer dans un tableau :
– sur ordinateur : cliquez sur une de ses cellules et naviguez horizontalement avec les touches de direction du clavier ou utilisez la barre de scroll horizontal qui se situe en bas du tableau.
– sur mobile : nous vous recommandons de passer en format paysage pour un meilleur affichage.
Vous pouvez télécharger la grille que vous souhaitez au format Excel tout en filtrant les mesureurs mais aussi la grille globale concernant la détection du trafic invalide au format PDF, avec les données des deux environnements renseignés.
Les mesureurs sont répartis dans plusieurs onglets selon si leur solution a été accréditée ou non par le MRC (Media Rating Council).
Environnement Web
Environnement In-app
Environnement Web
Solutions accréditées
 Updated April 16th, 2021 |  |  |  |  |  |  |
|---|---|---|---|---|---|---|
| 1) Do you flag invalid traffic on a user level? | Yes | Yes | Yes | Yes | No | Yes |
| 2) Do you flag invalid traffic on a site level? | Yes | Yes | Yes | Yes | Yes | Yes |
| 3) Do you flag invalid traffic on a domain level? | Yes | Yes | Yes | Yes | Yes | Yes and much more: Deterministic +Technical Evidence at the device level; HREF Domain; Platform Type; Provider; Site ID; Camapaign ID; Placement ID; Inventory Source; … |
| 4) What information is used for the IVT filter? (are they shared between campaigns, between clients?) | Adloox’s proprietary technologies, machine learning skills and the ability to process billions of queries per day worldwide are placing Adloox at the forefront of detection and elimination of Ad Fraud. Adloox analysis is user centric, URL centric, supply centric,.. hundreds of algorithms are applied every day to provide the most precise and granular data that can be optimized in PreBid. | DV has a dedicated fraud lab with data scientists, mathematicians and analysts from the cyber fraud prevention community. We employ a variety of approaches to detect new forms of fraud - from AI and machine learning to manual review. Through constant analysis, scenario management and research, we pinpoint the sites, apps and devices responsible for fraudulent activity - updating our protection in virtual real time. | IAS sees up to 10 billion impressions and trillions of events a day, from around the world. Their machine learning technology becomes smarter as it gets more data from advertisers and publishers. IAS clients don’t just benefit from the data they contribute but also from fraud detection results across the ecosystem. | IVT information is shared between campaigns and clients, so new learnings are immediately applied to all customers | Moat analyses all traffic on an impression-level basis | ALL HUMAN (Ex White Ops) collect device and network-level data subjected to hundreds of dynamic and continuously evolving HUMAN (Ex White Ops) verification challenges to identify evidence of compromise with greater accuracy than probabilistic or anomaly-based approaches. Payload Signals: * >2,000 signals collected in every execution * >10T interactions inspected every week Markers: Technical & Statistical Models >250 algorithms identifying threat model & bot category based on: * Technical evidence of automation * Behavioral patterns or other statistical anomalies |
| 5) Do you use Machine Learning methods for IVT filtering? | Yes | Yes | Yes | No | Yes | Yes but not only What we see: * 10 trillion interactions observed weekly * Billions of mobile devices observed weekly * Traffic from millions of domains What we do: * Reputation algorithms for all devices we’re observing * IP classification ML models: addresses, proxies, VPNs, mobile networks, residential, corporate, data center * Clustering algorithms of domains, binaries * Predictive ML engine that responds in <10 ms * Privacy & Compliance: Our privacy-sensitive code detects bots without tracking individual users across the internet (We have never and do not collect 3rd-party cookies) |
| 6) How often do you refresh the IVT filter parameters? (For instance, when an IP address is blacklisted, how long is it blacklisted for?) | Daily | DV is continually updating items and discovering and analyzing new fraudulent ADIDs throughout the day. Throughout the hour, these are pushed to DV’s systems | Once traffic is identified as IVT, IAS places the source utilized to identify the suspicious activity onto a blacklist. The IP blacklist is refreshed once a day to ensure accuracy, and new sources are added to it in near real time. | Parameters are refreshed after 12 hours | Daily | It is dynamic, real time and action per action |
| 7) Do you filter on auto-refresh impressions considering an excessive refresh rate as a risk of IVT? | Yes | Yes | No | No | Yes, daily | Yes, |
| 8) Do you partner with other companies to help you detect Invalid Traffic? | No | No | No | No | Yes, Digital Envoy | No |
| 9) Are you able to prevent ads from being viewed post-bid? | Yes | Yes | Yes | Yes | Yes | Yes |
| 10) Has your solution been accredited by MRC for Sophisticated Invalid Traffic? N.B: MRC Accreditation for General Invalid Traffic is included in the viewability accreditation | Yes | Yes | Yes | Yes | Yes | Yes, Desktop. Mobile, Connected TV |
| a) If yes, has your solution been continuously accredited since your first accreditation? | Yes | Yes | Yes | Yes | Yes | Yes |
| b) If no, are you in the process of being accredited by MRC? | N/A | N/A | N/A | N/A | N/A | N/A |
Environnement In-app
Solutions accréditées
Updated April 16th, 2021 | | | | |
|---|---|---|---|---|
| 1) Do you flag invalid traffic on a user level? | Yes | Yes | No | Yes |
| 2) Do you flag invalid traffic on a site level? | Yes (app level) | Yes (app level) | Yes (app level) | Yes |
| 3) Do you flag invalid traffic on a domain level? | Yes (app level) | Yes (app level) | Yes (app level) | Yes and much more: Deterministic +Technical Evidence at the device level; HREF Domain; Platform Type; Provider; Site ID; Camapaign ID; Placement ID; Inventory Source; … |
| 4) What information is used for the IVT filter? (are they shared between campaigns, between clients?) | DV has a dedicated fraud lab with data scientists, mathematicians and analysts from the cyber fraud prevention community. We employ a variety of approaches to detect new forms of fraud - from AI and machine learning to manual review. Through constant analysis, scenario management and research, we pinpoint the sites, apps and devices responsible for fraudulent activity - updating our protection in virtual real time. | IAS sees up to 10 billion impressions and trillions of events a day, from around the world. Their machine learning technology becomes smarter as it gets more data from advertisers and publishers. IAS clients don’t just benefit from the data they contribute but also from fraud detection results across the ecosystem. Additionally, IAS conducts offline technical research into apps and in-app inventory to identify suspicious activity. | Moat analyses all traffic on an impression-level basis | ALL HUMAN (Ex White Ops) collect device and network-level data subjected to hundreds of dynamic and continuously evolving HUMAN (Ex White Ops) verification challenges to identify evidence of compromise with greater accuracy than probabilistic or anomaly-based approaches. Payload Signals: * >2,000 signals collected in every execution * >10T interactions inspected every week Markers: Technical & Statistical Models >250 algorithms identifying threat model & bot category based on: * Technical evidence of automation * Behavioral patterns or other statistical anomalies |
| 5) Do you use Machine Learning methods for IVT filtering? | Yes | Yes | Yes | Yes but not only What we see: * 10 trillion interactions observed weekly * Billions of mobile devices observed weekly * Traffic from millions of domains What we do: * Reputation algorithms for all devices we’re observing * IP classification ML models: addresses, proxies, VPNs, mobile networks, residential, corporate, data center * Clustering algorithms of domains, binaries * Predictive ML engine that responds in <10 ms * Privacy & Compliance: Our privacy-sensitive code detects bots without tracking individual users across the internet (We have never and do not collect 3rd-party cookies) |
| 6) How often do you refresh the IVT filter parameters? (For instance, when an IP address is blacklisted, how long is it blacklisted for?) | DV is continually updating items and discovering and analyzing new fraudulent ADIDs throughout the day. Throughout the hour, these are pushed to DV’s systems | Once traffic is identified as IVT, IAS places the source utilized to identify the suspicious activity onto a blacklist. The IP blacklist is refreshed once a day to ensure accuracy, and new sources are added to it in near real time. | Daily | It is dynamic, real time and action per action |
| 7) Do you filter on auto-refresh impressions considering an excessive refresh rate as a risk of IVT? | Yes | No | Yes, daily | Yes |
| 8) Do you partner with other companies to help you detect Invalid Traffic? | No | No | Yes, Digital Envoy | No |
| 9) Are you able to prevent ads from being viewed post-bid? | Yes | Yes | No, Blocking is not available currently on the app level. | Yes |
| 10) Has your solution been accredited by MRC for Sophisticated Invalid Traffic? N.B: MRC Accreditation for General Invalid Traffic is included in the viewability accreditation | Yes, for mobile/tablet in-app as well as the CTV environment | Yes | Yes | Yes, Desktop. Mobile, Connected TV |
| a) If yes, has your solution been continuously accredited since your first accreditation? | Yes | Yes | Yes | Yes |
| b) If no, are you in the process of being accredited by MRC? | N/A | N/A | N/A | N/A |
Solutions non accréditées
Updated April 16th, 2021 | | |
|---|---|---|
| 1) Do you flag invalid traffic on a user level? | Yes | Yes |
| 2) Do you flag invalid traffic on a site level? | Yes (app level) | Yes (app level) |
| 3) Do you flag invalid traffic on a domain level? | Yes (app level) | Yes (app level) |
| 4) What information is used for the IVT filter? (are they shared between campaigns, between clients?) | Adloox’s proprietary technologies, machine learning skills and the ability to process billions of queries per day worldwide are placing Adloox at the forefront of detection and elimination of Ad Fraud. Adloox analysis is user centric, URL centric, supply centric,.. hundreds of algorithms are applied every day to provide the most precise and granular data that can be optimized in PreBid. | IVT information is shared between campaigns and clients, so new learnings are immediately applied to all customers |
| 5) Do you use Machine Learning methods for IVT filtering? | Yes | No |
| 6) How often do you refresh the IVT filter parameters? (For instance, when an IP address is blacklisted, how long is it blacklisted for?) | Daily | Parameters are refreshed after 12 hours |
| 7) Do you filter on auto-refresh impressions considering an excessive refresh rate as a risk of IVT? | Yes | No |
| 8) Do you partner with other companies to help you detect Invalid Traffic? | No | No |
| 9) Are you able to prevent ads from being viewed post-bid? | Yes | Yes |
| 10) Has your solution been accredited by MRC for Sophisticated Invalid Traffic? N.B: MRC Accreditation for General Invalid Traffic is included in the viewability accreditation | Under review | No |
| a) If yes, has your solution been continuously accredited since your first accreditation? | N/A | N/A |
| b) If no, are you in the process of being accredited by MRC? | Under review | No |