SIVT is for Sophisticated Invalid Traffic, while GIVT stands for General Invalid Traffic. It is defined by the MRC (Media Ratings Council). It avoids using broad terminology like fraud and thus classifies traffic as "invalid."
According to MRC, GIVT is traffic recognized using systematic filtration methods such as the use of lists or other standardised parameter checks. "Data Center Traffic," "Bots, Spiders, and Crawlers," "Activity-based filtration," "Non-browser UA Headers or other unknown browsers," and "pre-fetch or browser pre-rendered traffic" are a few instances of GIVT traffic.
In a server-to-server communication, particularly in real-time bidding. SSPs/DSPs must pre-validate ad request IPs and filter such traffic if it originates from any Data Centers or VPNs. There are billions of IPs associated with data centers, VPNs, and proxy servers, and the number is growing. Simple to recognise, but difficult to filter in real-time because ad response time is a big limitation.
SIVT, according to MRC, is comprised of more difficult-to-detect circumstances that necessitate advanced analytics, multi-point corroboration/coordination, significant human engagement, and so on to evaluate and identify. "Differentiating human and IVT if they are originating from the same source," "bots masquerading as legitimate users," "Hijacked user sessions, devices, etc.," "Non-viewable / obfuscated ad serving," "Invalid Proxy Traffic," "Adware and Malware," "Incentivized traffic," "Falsified viewable impressions," "Falsified sites," "Cookie stuffing, harvesting and recycling,"
According to our CTO (Anand Kumar)
We are witnessing 30-40% of traffic that is not created by human devices, and in most cases, sources are doing it on purpose to benefit. Amli filters billions of IPs and evaluates their percentage for real-time auto-blocking. Caching and managing such large IP addresses was a difficult challenge at first, but we managed it without utilising separate servers and with 100 percent uptime. It takes us less than 1 milliseconds to filter, with no additional strain on equipment. In the current circumstances, it appears positive for us because our ad response time is still less than 80 milliseconds.
Post validations are important components in filtering and eliminating SIVT sources; they begin with validation against request data and beacons (for example, impression and click) data that we receive directly from the users' devices.
Ad networks should avoid using direct payload and instead use their own encoding techniques that can only be decoded on their ad servers to avoid exposing any entities.
Following that, we validate the beacons headers data against the request data. It includes User-Agent, IP, Bundle, and Domain, among other things. Finally, our machine learning algorithms track the legitimate behaviour of others. Ad viewability is another element that we filter and exclude altogether or limit such sources where ad viewability is less than 1 second with the proper MRC defined pixel. Our IVT filtration technology is the greatest, and I can guarantee it.