Skynet, the software that identifies the target “terrorists” to kill – BFMTV.COM

NSA agents (National Security Agency) have called Skynet, as in the saga Terminator . C e-secret US computer program’s mission is to identify terrorist targets in Pakistan, which can then be eliminated by the American services, based on metadata. Including under a form of preventive self-defense.

The existence of this system was revealed in April 2015 by the magazine The Intercept which was based on documents provided by Edward Snowden . In its survey published in the Wednesday edition, Le Monde said that Michael Hayden, former head of the NOS and CIA are had hinted at an academic conference:

“Yes, we kill people Based on metadata,” he told

<. h3> Software “Big data” rather “ordinary”

The operation of Skynet is that “an ordinary application of Big Data” advance Le Monde , based on documents provided by Edward Snowden. At the end of the process, the Pakistani population is classified very binary fashion into two categories: “innocent” or “terrorists”. These are then removed by US drone strikes.

So how Skynet he distributes the right to life or the act of death? Everything starts from the analysis of the famous metadata drawn off to mobile operators. It is not eavesdropping of messages and communications as such. Here it is not so much the content of the communications is important for the sender, receiver, transmission time, date, size … Not to mention the location of the protagonists, driven by the widespread use mobile phones. And, said the newspaper, it is useless to change SIM card ( IMEI number attached to the phone will do), or even to exchange the equipment. The software seamlessly integrates this kind of subterfuge.

“margin of error”

Once the data collected, it is stored and analyzed by Skynet who has this feature to learn by itself. To discriminate the “terrorists”, the program is also based on a “ground truth”. It is a “set of data discussed and annotated in which the mobile phone users have previously been classified” in the above two categories.

After sorting, the NSA -in light provides a “margin of error” – the list of targets to eliminate. The final decision is then up to the State Department or the CIA.

Potentially 15,000 false positives

Acting as a service provider, NSA boasts surgical precision. False positives, ie those identified as “terrorists” as they are “innocent” would be in the best case, “only” 0.008%. But compared to the population of Pakistan is 189 million people, this is still some 15,000 people, said the evening news. Paradoxically, 50% of the “terrorists” are not included because their “score” dangerousness established by the NSA algorithm remains below a threshold.

The method is, effective? This is the problem of Skynet. According to the Bureau of Investigative Journalism (BIJ) based in London, which holds the accounts of the US strikes in Pakistan, 421 drone attacks were carried out by the CIA in 2014 and 2015. The most coins minted often at night, on civilian homes. According to the BIJ, 22% of those killed were civilians, and only 1.5% of target “high value”. The approximately 76% of other deaths would however not be classified into one or the other of these two categories.