Unique Impressions is a Carnivore client, a network visualization and sonification software program that measures the amount communication activity between computers on the local network, and between computers on the local network and the Internet. More specifically, Unique Impressions measures the amount of activity in each “unique impression” on a network.

Overview
The core of the Unique Impressions application is contained in a Java object which listens for network activity and generates commands to fill a matrix of pixels. First, a sender/requester IP pair is obtained by the software. If this IP pair has not been previously encountered, then it is considered a unique impression. Visuals and sounds are generated according to these properties.

Visualization Details
Currently, two different visualizations are generated by the system. The first involves adding a pixel for each unique impression and changing the opacity of the pixels already added. Every incoming packet is analyzed. If it is determined to be unique, then a pixel is added at full opacity. The color of the pixel is derived from the ip numbers of the sender and receiver. The first three values of the IP address are mapped to rgb values Each of the corresponding rgb values for both IPs are then “mixed” to arrive at one color for that connection. If it has been previously encountered then that unique impression’s pixel is already in the matrix. In this case, the system adds to the opacity of that pixel. The system then continuously monitors incoming packets, adding pixels for new connections, increasing opacity for active old connections and decreasing opacity for inactive old connections. The second visualization is an extension of the first where the rgb values obtained from the IP numbers are mapped to x/y/z coordinates in 3-D space. In this case every unique impression becomes a primitive (such as a line or triangle) in a specific location in 3-D space. The opacity is mapped to an envelope follower which grows or shrinks the shape primitive based on the amount of activity present at that connection. Thus active connections grow and inactive ones shrink.

Sonification Details
While several sonification techniques have been attempted thus far, including granular synthesis, oscillator banks and wavetables, the only method currently being implemented involves reading the matrix of pixels as an audio signal. The matrix of pixels is made up of four planes: red, green, blue and alpha. Each plane of each pixel in the matrix is read, line by line. These values are then output as an audio signal. In addition, the red, green and blue values are averaged, with the resulting value being used to alter the cutoff frequency of an audio filter. The more network activity, the higher the cutoff frequency. The red, green and blue signals are run through this filter. Thus, similar to the way the 3-D primitives grow and shrink based on network activity, here the signal gets more and more filtered (and thus less audible) as overall network activity decreases

Unique Impressions seeks to explore the possibility that if the Internet can be seen to be a self-organizing organic entity, then can the information in it be thought of as its own entity and not a reference to entities in the physical world? Instead of a visualization representing data symbolically, by filtering it through known or recognizable visual forms which mimic some aspect of physical reality, the data can simply represent itself. So the visualization isn’t a representation of reality but a piece or a residue of reality (and thus *is* reality in a sense). The sounds and visuals produced by Unique Impressions will undoubtedly resemble noise or random patterns. However, if we treat this “noisy” data as “reality”(remembering that we’re operating on the premise that the Internet is a self-organizing organic system), and do not represent it symbolically but let it represent itself indexically, then perhaps certain relationships will appear, or patterns will emerge that would not otherwise manifest themselves.

The title of the piece derives from the concept of “unique impressions”, often used by Internet marketers and advertisers to catalog and track to the amount of visitors to a site or the amount of times an advertising banner has been clicked. The idea of a unique impression is used to discriminate between multiple/unique users visiting a site versus a single user that generates many “hits”. Thus, a single user can generate many hits to a web site but only one unique impression – “uniqueness” being determined by the IP address of the document requester.

GitHub