Le Monde | • Updated | By
Scientists at the prestigious National Institute for Computer Science and Control (INRIA) are more than skeptical about the law on intelligence, adopted at first reading on 5 May Assembly
Read also:. The bill overwhelmingly approved the information to the Assembly
In a note dated April 30, which was originally to be sent in ministerial offices and deputies and Le Monde has obtained, INRIA Weeder certain technical and controversial points of the bill.
No anonymous data
The Institute, under the supervision of Bercy and the Ministry of Research, sharply critical article 851-4 of the law, which provides for the establishment of algorithms – the famous “black boxes” – to automatically detect terrorist behavior on the Internet. The law provides that the analysis is done on anonymous data, identifying any intervening only when a threat is detected: an argument hammered by the government during consideration of the text to the Assembly and the first hearings in the Senate. “There is no safe anonymization technique today. A law should not be based on the notion of anonymous or anonymized data “ INRIA slice.
Computer National Commission and Freedom had already pointed the little anonymous data that can, if necessary, identify a suspect
Read also:. Critics of the CNIL against the bill on intelligence
The dangers of
The Institute then alerts on potential excesses of algorithmic detection of terrorists. A computer program, however well regulated, consistently produces errors, which are more numerous than the mass of data to be processed is large, the scientists explain in their notes.
“This phenomenon well-known scientist (…) has implications that the text of the bill does not take into account “, warns INRIA
Read also:. Law intelligence “wrongly Tens of thousands of people will be suspected”
INRIA also noted the ineffectiveness of digital surveillance introduced by algorithms, “easily circumvented even without technical knowledge developed “. The public institute says that when these circumvention devices – such as a VPN (the user will connect, securely, to a third computer which handles its internet navigation) – will be implemented, “no information on the final destination of the information or the message content will be possible “.
Finally, Inria, held ” available to the legislator, “ believes that the composition of the national intelligence oversight technical Commission (CNCTR) responsible for auditing the plays, and thus algorithms, is not satisfactory. “In light of scientific and technical complexity of digital topics” , it should benefit the Institute wrote “balanced representation between digital and legal expertise” . Under current law, only one member of the CNCTR, appointed by the Regulatory Authority for electronic communications and postal services, will have a technical background allows him to control the algorithm.
A second note in preparation
The position expressed in this note is not a surprise and largely reflects the various interventions in the media, the scientific members of the institute.
Why this document, very rare on the part of the Institute, finally he was not sent to MPs and ministers? Dated 30 April, it was completed after the debates in the Assembly session and just before the solemn vote: too late to bring about changes in the text. At INRIA, however, this publication is not justified by the desire to adopt “constructive approach” and not limited to a simple “statement” .
The institution’s reluctance to publish its official opinion also reflects the difficulty for the public body under ministerial supervision to take a position frontally opposed to the choices of the government.
According to our information, a second note with concrete proposals in preparation. It will give the key to the government to correct the defects that the Institute notes from his first note
.
No comments:
Post a Comment