August 2021 - nexa - server-nexa.polito.it

Facebook is obstructing our work on disinformation. Other researchers could be next | Laura Edelson and Damon McCoy | The Guardian

by Alberto Cammozzo

<https://www.theguardian.com/technology/2021/aug/14/facebook-research-disinf…> Last week, Facebook disabled our personal accounts, obstructing the research we lead at New York University to study the spread of disinformation on the company’s platform. The move has already compromised our work – forcing us to suspend our investigations into Facebook’s role in amplifying vaccine misinformation, sowing distrust in our elections and fomenting the violent riots at the US Capitol on 6 January. But even more important than the effect on our work is what Facebook’s hostility toward outside scrutiny means for the many other researchers and journalists trying to study Facebook’s effects on society. We’ve already heard from other researchers planning similar projects who are now pulling back. If Facebook has its way, there will be no independent research of its platform. Our dispute with Facebook centers on a research tool called Ad Observer. Ad Observer is a web browser extension that Facebook users can choose to install to share with us limited and anonymous information about the ads that Facebook shows them. The data they share with us includes the categories advertisers chose when targeting them. Examples might be “married women” or “interested in dating” or “lives in Florida”. Using data collected through Ad Observer, and also data collected using the transparency tools Facebook makes available to researchers, we’ve been able to help the public understand how the platform fails to live up to its promises, and shed light on how it sells the attention of its users to advertisers. In a forthcoming paper, we show that Facebook has failed to include more than 100,000 ads that meet its own criteria as political, social and issue ads in its public archive. For example, it failed to include ads supporting Joe Biden ahead of the 2020 elections; Amazon ads about the minimum wage; and an anti-mask ad targeted to conservatives run by a group called Reopen USA, whose Facebook page posts anti-vaccine and anti-mask memes. We have also shown how highly partisan and misleading news sources get far more engagement on Facebook than reliable news sources do, and we will be publishing an expanded version of this analysis in another forthcoming paper. But we’re not the only researchers who use Ad Observer’s data. For the past three years, we’ve been making information we collect through Ad Observer and through Facebook’s tools available to other researchers and journalists, so they can conduct their own investigations. The Markup has used our data to report on how ads with QAnon content and merchandise from extremist militia groups have slipped through Facebook’s filters, despite bans. The Markup also used the data to demonstrate how corporate advertisers such as ExxonMobil and Comcast promote seemingly contradictory messages about hot button issues to different audiences. Reporters from Florida to Kentucky to New Mexico used it to report on trends in political advertising in their states ahead of the 2020 elections. In disabling our accounts last week, Facebook claimed that we were violating its terms of service, that we were compromising user privacy, and that it had no choice but to shut us down because of an agreement it has with the Federal Trade Commission. All of these claims are wrong. Ad Observer collects information only about advertisers, not about our volunteers or their friends, and the FTC has stated that our research does not violate its consent decree with Facebook. Unfortunately, Facebook’s campaign against us is part of a larger pattern of hostility toward outside scrutiny. Just last month the New York Times reported that Facebook, after internal controversy, was dismantling a team working on CrowdTangle, its marquee transparency tool for researchers who want to see how unpaid Facebook posts spread and gain engagement on the platform. The paper reported this week that the White House itself was having so much trouble getting a straight answer from Facebook on vaccine misinformation that officials asked to speak directly to the platform’s data scientists, rather than its lobbyists. Social Science One, launched in 2018 to great fanfare, was supposed to provide researchers with access to Facebook user data in a safe way. But the data offered proved to be much less useful than anticipated, to the point that the funders of the project dropped out. In our work we’ve shown how Facebook’s transparency tools fall short of promises. We can’t let Facebook decide unilaterally who gets to study the company and what tools they can use. The stakes are too high. What happens on Facebook affects public trust in our elections, the course of the pandemic and the nature of social movements. We need the greater understanding that researchers, journalists and public scrutiny provide. If Facebook won’t allow this access voluntarily, then it’s time for lawmakers to require it.

Aug. 15, 2021

I sindacati e l’arrocco al tempo della rivoluzione digitale

by don Luca Peyron

Un editoriale duro. Al di là del tema singolo mi pare oggettivo il fatto che un ente intermedio così importante come il sindacato, sui temi che ci riguardano in questa lista, non sembra particolarmente presente o propositivo. Ne ho personalmente riprova nel percorso fatto con il territorio in occasione della candidatura di Torino quale sede del Centro Italiano per l'Intelligenza Artificiale. Buona lettura. dl https://www.corrierecomunicazioni.it/lavoro-carriere/competenze/i-sindacati… _________________________ don Luca Peyron Direttore Ufficio per la Pastorale Universitaria Arcidiocesi di Torino www.universitari.to.it via XX settembre 83, Torino tel. 011 5156239

Aug. 14, 2021

Non solo cyber, così riformiamo l’intelligence. Parla Franco Gabrielli

by don Luca Peyron

Una intervista, non un paper, ma comunque capace di fornire indicazioni utili alla nostra riflessione, penso. Buona lettura dl https://formiche.net/2021/08/non-solo-cyber-cosi-riformiamo-lintelligence-p… _________________________ don Luca Peyron Direttore Ufficio per la Pastorale Universitaria Arcidiocesi di Torino www.universitari.to.it via XX settembre 83, Torino tel. 011 5156239

Aug. 14, 2021

Philip E. Agre

by J.C. DE MARTIN

Confesso che fino ad oggi avevo mai sentito parlare di Philip E. Agre, informatico diventato umanista (a UCLA!): https://www.washingtonpost.com/technology/2021/08/12/philip-agre-ai-disappe… https://en.wikipedia.org/wiki/Philip_Agre Qui molte sue pubblicazioni: https://pages.gseis.ucla.edu/faculty/agre/ A giudicare dai titoli molte di esse mi sembrano potenzialmente d'interesse per questa lista. jc

Aug. 13, 2021

Young People Have More Faith in Algorithms than Politicians

by don Luca Peyron

Young people champion an open internet but are concerned about misuse young people would be more likely to trust a system run by artificial intelligence than humans The Youth Recovery Plan is an initiative of the World Economic Forum Global Shapers Community, a network created to ensure that the next generation – people between the ages of 20 and 30 – are empowered to help shape our common future. They are active in more than 450 cities in 150 countries world, and together they implement local and global projects to improve the state of the world. Considerazioni quanto mai interessanti, buona lettura dl https://www.weforum.org/press/2021/08/young-people-have-more-faith-in-algor… _________________________ don Luca Peyron Direttore Ufficio per la Pastorale Universitaria Arcidiocesi di Torino www.universitari.to.it via XX settembre 83, Torino tel. 011 5156239

Aug. 12, 2021

Researchers Create 'Master Faces' to Bypass Facial Recognition

by Alberto Cammozzo

<https://www.vice.com/en/article/k78ygn/researchers-create-master-faces-to-b…> Researchers Create 'Master Faces' to Bypass Facial Recognition According to the paper, their findings imply that facial recognition systems are “extremely vulnerable.” Researchers have demonstrated a method to create "master faces," computer generated faces that act like master keys for facial recognition systems, and can impersonate several identities with what the researchers claim is a high probability of success. In their paper <https://arxiv.org/pdf/2108.01077.pdf>, researchers at the Blavatnik School of Computer Science and the School of Electrical Engineering in Tel Aviv detail how they successfully created nine "master key" faces that are able to impersonate almost half the faces in a dataset of three leading face recognition systems. The researchers say their results show these master faces can successfully impersonate over 40 percent of the population in these systems without any additional information or data of the person they are identifying. The paper cites previous research which showed a similar method for creating master fingerprints <https://ieeexplore.ieee.org/document/7893784>. According to the paper, their findings imply that facial recognition systems are “extremely vulnerable.” The "master key" faces tended to be older, and didn't have glasses or facial hair. The researchers tested their methods against three deep face recognition systems–Dlib, FaceNet, and SphereFace. Lead author Ron Shmelkin told Motherboard that they used these systems because they are capable of recognizing “high-level semantic features” of the faces that are more sophisticated than just skin color or lighting effects. The researchers used a StyleGAN to generate the faces and then used an evolutionary algorithm and neural network to optimize and predict their success. The evolutionary strategy then creates iterations, or generations, of candidates of varying success rates. The researchers then used the algorithm to train a neural network, to classify the best candidates as the most promising ones. This is what teaches it to predict candidates’ success and, in turn, direct the algorithm to generate better candidates with a higher probability of passing. “We are interested in further exploring the possibility of using the master faces generated by our method in order to help protect existing facial recognition systems from such attacks,” Shmelkin told Motherboard. The researchers even predict that their master faces could be animated using deepfake technology to bypass liveness detection, which is used to determine whether a biometric sample is real or fake. The paper also notes that white males over the age of 60 in the University of Massachusetts’ Labeled Faces in the Wild (LFW) dataset tended to be less varied compared to younger groups, so much of that group could be covered by a single older master face. Additionally, only two of the nine master faces created were female, which the paper notes matches the “much lower frequency” of female faces in the LFW dataset (22 percent.) The success of their findings shows how facial recognition software can be flawed and biased. Its continued use by law enforcement has resulted in multiple wrongful arrests <https://www.vice.com/en/article/xgx5gd/man-wrongfully-arrested-by-facial-re…>, even when it has proven to be easy to tamper with <https://www.vice.com/en/article/m7e9qv/hackers-fool-facial-recognition-into…>.

Aug. 12, 2021

EvilModel: Hiding Malware Inside of Neural Network Models

by 380°

Buongiorno, mi sono sempre chiesto /se/, essendo «poorly explainable» (assimilabili a codice binario o solo troppo complessi?), i modelli di machine learning non fossero un vettore di infezione ideale. «EvilModel: Hiding Malware Inside of Neural Network Models» Submitted on 19 Jul 2021 (v1), last revised 5 Aug 2021 (this version, v4) --8<---------------cut here---------------start------------->8--- Abstract: Delivering malware covertly and evasively is critical to advanced malware campaigns. In this paper, we present a new method to covertly and evasively deliver malware through a neural network model. Neural network models are poorly explainable and have a good generalization ability. By embedding malware in neurons, the malware can be delivered covertly, with minor or no impact on the performance of neural network. Meanwhile, because the structure of the neural network model remains unchanged, it can pass the security scan of antivirus engines. Experiments show that 36.9MB of malware can be embedded in a 178MB-AlexNet model within 1% accuracy loss, and no suspicion is raised by anti-virus engines in VirusTotal, which verifies the feasibility of this method. With the widespread application of artificial intelligence, utilizing neural networks for attacks becomes a forwarding trend. We hope this work can provide a reference scenario for the defense on neural network-assisted attacks. --8<---------------cut here---------------end--------------->8--- Non ho letto il paper, solo l'abstract, e capisco perché un semplice antivirus (porello) non è in grado di distinguere il virus dagli altri neuroni; una possibile soluzione non potrebbe essere quella di fornire anche un hash (chessò un sha512) che garantisca che il modello non sia stato alterato? Oltre questo, non /dovrebbe/ comunque essere messa a disposizione tutta la catena di build e gli input per poter riprodurre il modello bit-per-bit, ala "reproducible build"? Oltre a questo, l'intero sistema generato con tecniche ML non dovrebbe essere completamente bootstrappable? Esagero? Saluti, 380° -- 380° (Giovanni Biscuolo public alter ego) «Noi, incompetenti come siamo, non abbiamo alcun titolo per suggerire alcunché» Disinformation flourishes because many people care deeply about injustice but very few check the facts. Ask me about <https://stallmansupport.org>.

Aug. 10, 2021

Re: [nexa] Ma che davero?!? (was Re: Virus laziale)

by Giacomo Tesio

Ti sbagli Giovanni, On Tue, 10 Aug 2021 15:06:32 +0200 380° wrote: > > Chi scrive di queste cose NON PUÓ ignorare che ... > NON PUÓ ignorare che ... > NON PUÓ ignorare che ... Mi pare evidente che può. Non DOVREBBE, ma tanto l'etica nell'informatica è un problema delle intelligenze artificiali, non riguarda mica gli "esperti"! Ed infatti, qualunque sciocchezza scrivano [1] viene pubblicata. Il passaggio più qualunquista e superficiale del pezzo secondo me è: > [...] Nel caso in questione non sarà così utile trovare un colpevole o > anche solo un capro espiatorio; Mi piacerebbe chiedere a Mantellini a cosa serva pagare responsabili IT se tanto non devono rispondere delle loro decisioni. Fintanto che i responsabili non rispondono perché "tanto è inutile cercare un capro espiatorio", i "responsabili" più incompetenti godono di un enorme vantaggio competitivo rispetto a quelli competenti. Perché quelli competenti spesso dicono "No" ai superiori, mentre quelli incompetenti dicono sempre "Sì", consapevoli che tanto non risponderanno delle conseguenze. E così ai vertici delle infrastrutture IT ci si ritrova pieni di incompetenti ben retribuiti a scegliere sistematicamente la soluzione tecnicamente peggiore per l'azienda/PA. Un incidente come quello avvenuto in Regione Lazio dovrebbe comportare l'automatico licenziamento di tutta la catena dirigenziale, con tanto di causa per danno erariale. Perché sono "responsabili" evidentemente irresponsabili di ciò per cui sono pagati per rispondere. Non ci dovrebbero essere scuse. E' difficile? Imparino! E costoso? Riprogettino! Invece, per Mantellini, la responsabilità è inutile. ¯\_(ツ)_/¯ Mantellini conclude con uno straordinario momento di lucidità: > Nel lungo periodo, come in tutte le cose della nostra vita, ci > salveranno solo la scuola e la cultura, e quindi quanto fin da ora > sapremo fare per avere nuove generazioni più curiose e diverse da noi. > > Ci servirà formare cittadini del mondo che siano migliori di noi, che > sappiano più cose di noi E' però un attimo fugace: > Ci serve una nuova Italia di persone che, a vent’anni come a ottanta, > scelgano di smettere di guardare disperatamente indietro. Come se l'eccessiva memoria o la cultura storica soverchiante fossero un problema di questo Paese! Dannati storici! :-D Noi abbiamo un drammatico bisogno di persone che abbiano il coraggio di guardare con onestà intellettuale al passato per riconoscerne ed evitare di ripeterne continuamente gli errori. L'ansia del futuro non è migliore della nostalgia del passato. Perché ogni visione del futuro, per essere realizzabile, deve essere costruita sempre sulla comprensione del passato. Giacomo [1] Uno sproloquio fra tanti, incidentalmente dello stesso autore, https://www.internazionale.it/opinione/massimo-mantellini/2020/06/05/immuni…

Aug. 10, 2021

Climate Change 2021: The Physical Science Basis

by don Luca Peyron

Una tecnologia per il pianeta, questa certamente un'urgenza. Quanto ci preoccupiamo di una tecnica che risolva problemi tecnici, come il ponte sullo stretto, ed intanto si muore ancora di fame? Che il clima torrido scaldi anche il cuore e riattivi sinapsi sociali addormentate da un sistema tecnico che ha alimentato anche troppo se stesso. Buona lettura dl https://www.ipcc.ch/report/sixth-assessment-report-working-group-i/ _________________________ don Luca Peyron Direttore Ufficio per la Pastorale Universitaria Arcidiocesi di Torino www.universitari.to.it via XX settembre 83, Torino tel. 011 5156239

Aug. 10, 2021

Riforma diritto d’autore ed equo compenso: cosa cambia in Italia per gli editori

by don Luca Peyron

Cominciato l'iter di ricezione della normativa europea, un primo sunto Buona lettura dl https://www.money.it/Riforma-diritto-autore-equo-compenso-cosa-cambia-Itali… _________________________ don Luca Peyron Direttore Ufficio per la Pastorale Universitaria Arcidiocesi di Torino www.universitari.to.it via XX settembre 83, Torino tel. 011 5156239

Aug. 10, 2021