Comprendre informations du langage, algorithms

Making algorithms understand what we are talking about

/0 Comments/in , /by

Human language contains different types of information. We understand it all unconsciously, but explaining it systematically is much more difficult. The same is true for machines. The NoRDF Project Chair “Modeling and Extracting Complex Information from Natural Language Text” seeks to solve this problem: how can we teach algorithms to model and extract complex information from language? Fabian Suchaneck and Chloé Clavel, both researchers at Telecom Paris, explain the approaches of this new project

What aspects of language are involved in making machines understand?

Fabian Suchaneck: We need to make them understand more complicated natural language texts. Current systems can understand simple statements. For example, the sentence: “A vaccine against Covid-19 has been developed” is simple enough to be understood by algorithms. On the other hand, they cannot understand sentences that go beyond a single statement, such as: “If the vaccine is distributed, the Covid-19 epidemic will end in 2021. In this case, the machine does not understand that the condition required for the Covid-19 epidemic to end in 2021 is that the vaccine is distributed. We also need to make machines understand what emotions and feelings are associated with language; this is Chloé Clavel’s specialist area.

What are the preferred approaches in making algorithms understand natural language?

FS: We are developing “neurosymbolic” approaches, which seek to combine symbolic approaches with deep learning approaches. Symbolic approaches use human-implemented logical rules that simulate human reasoning. For the type of data we process, it is fundamental to be able to interpret what has been understood by the machine afterwards. Deep learning is a type of automatic learning where the machine is able to learn by itself. This allows for greater flexibility in handling variable data and the ability to integrate more layers of reasoning.

Where does the data you analyze come from?

FS: We can collect data when humans interact with chatbots from a company and especially those from the project’s partner companies. We can extract data from comments on web pages, forums and social networks.

Chloé Clavel: We can also extract information about feelings, emotions, social attitudes, especially in dialogues between humans or humans with machines.

Read on I’MTech: Robots teaching assistants

What are the main difficulties for the machine in learning to process language?

CC: We have to create models that are robust in changing contexts and situations. For example, there may be language variability in the expression of feelings from one individual to another, meaning that the same feelings may be expressed in very different words depending on the person. There is also a variability of contexts to be taken into account. For example, when humans interact with a virtual agent, they will not behave in the same way as with a human, so it is difficult to compare data from these different sources of interactions. Yet, if we want to move towards more fluid and natural human-agent interactions, we must draw inspiration from the interactions between humans.

How do you know whether the machine is correctly analyzing the emotions associated with a statement?

CC: The majority of the methods we use are supervised. The data entered into the models are annotated in the most objective way possible by humans. The goal is to ask several annotators to annotate the emotion they perceive in a text, as the perception of an emotion can be very subjective. The model is then taught about the data for which a consensus among the annotators could be found. When testing the performance of the model, when we inject an annotated text into a model that has been trained with similar texts, we can see if the annotation it produces is close to those determined by humans.

Since the annotation of emotions is particularly subjective, it is important to determine how the model actually understood the emotions and feelings present in the text. There are many biases in the representativeness of the data that can interfere with the model and mislead us on the interpretation made by the machine. For example, if we assume that younger people are angrier than older people in our data and that these two categories do not express themselves in the same way, then it is possible that the model may end up simply detecting the age of the individuals and not the anger associated with the comments.

Is it possible that the algorithms end up adapting their speech according to perceived emotions?

CC: Research is being conducted on this aspect. Chatbots’ algorithms must be relevant in solving the problems they are asked to solve, but they must also be able to provide a socially relevant response (e.g. to the user’s frustration or dissatisfaction). These developments will improve a range of applications, from customer relations to educational or support robots.

What contemporary social issues are associated with the understanding of human language by machines?

FS: This would notably allow a better understanding of the perception of news on social media by humans, the functioning of fake news, and therefore in general which social group is sensitive to which type of discourse and why. The underlying reasons why different individuals adhere to different types of discourse are still poorly understood today. In addition to the emotional aspect, there are different ways of thinking that are built in argumentative bubbles that do not communicate with each other.

In order to be able to automate the understanding of human language and exploit the numerous data associated with it, it is therefore important to take as many dimensions into account as possible, such as the purely logical aspect of what is said in sentences and the analysis of the emotions and feelings that accompany them.

By Antonin Counillon

IoT, Internet of Things

A standardized protocol to respond to the challenges of the IoT

/0 Comments/in , /by

The arrival of 5G has put the Internet of Things back in the spotlight, with the promise of an influx of connected objects in both the professional and private spheres. However, before witnessing the projected revolution, several obstacles remain. This is precisely what researchers at IMT Atlantique are working on, and they have already achieved results of global significance.

The Internet of Things (IoT) refers to the interconnection of various physical devices via the Internet for the purpose of sharing data. Sometimes referred to as the “Web 3.0”, this field is set to develop rapidly in the coming years, thanks to the arrival of new networks, such as 5G, and the proliferation of connected objects. Its applications are infinite: monitoring of health data, the connected home, autonomous cars, real-time and predictive maintenance on industrial devices, and more.

Although it is booming, the IoT still faces major challenges. “We need to respond to three main constraints: energy efficiency, interoperability and security,” explains Laurent Toutain, a researcher at IMT Atlantic. But there is one problem: these three aspects can be difficult to combine.

The three pillars of the IoT

First, energy is a key issue for the IoT. For most connected objects, the autonomy of a smartphone is not sufficient. In the future, a household may have several dozen such devices. If they each need to be recharged every two or three days, the user will have to devote several hours to this task. And what about factories that could be equipped with thousands of connected objects? In some cases, these are only of value if they have a long battery life. For example, a sensor could be used to monitor the presence of a fire extinguisher at its location and send an alert if it does not detect one. If you have to recharge its battery regularly, such an installation is no longer useful.

For a connected object, communication features account for the largest share of energy consumption. Thus, the development of IoT has been made possible by the implementation of networks, such as LoRa or Sigfox, allowing data to be sent while consuming little energy.

The second issue is interoperability, i.e. the ability of a product to work with other objects and systems, both current and future. Today, many manufacturers still rely on proprietary universes, which necessarily limits the functionalities offered by the IoT. Take the example of a user who has bought connected light bulbs from two different brands. They will not be able to control them via a single application.

Finally, the notion of security remains paramount within any connected system. This observation is all the more valid in the IoT, especially with applications involving the exchange of sensitive data, such as in the health sector. There are indeed many risks. An ill-intentioned user could intercept data during transmission, or send false information to connected objects, thus inducing wrong instructions, with potentially disastrous consequences.

Read more on I’MTech: The IoT needs dedicated security – now

On the Internet, methods are already in place to limit these threats. The most common is end-to-end data encryption. Its purpose is to make information unreadable while it is being transported, since the content can only be deciphered by the sender and receiver of the message.

Three contradictory requirements?

Unfortunately, each of the three characteristics can influence the others. For example, by multiplying the number of possible interlocutors, interoperability raises more security issues. But it also affects energy consumption. “Today, the Internet is a model of interoperability,” explains Laurent Toutain. For this, it is necessary to send a large amount of information each time, with a high degree of redundancy. It offers remarkable flexibility, but it also takes up a lot of space.” This is only a minor disadvantage for a broadband network, but not for the IoT, which is constrained in its energy consumption.

Similarly, if you want to have a secure system, there are two main possibilities. The first is to close it off from the rest of the ecosystem, in order to reduce risks, which radically limits interoperability.

The second is to implement security measures, such as end-to-end encryption, which results in more data being sent, and therefore increased energy consumption.

Reducing the amount of data sent, without compromising security

For about seven years, Laurent Toutain and his teams have been working to reconcile these different constraints, in the context of the IoT. “The idea is to build on what makes the current Internet so successful and adapt it to the constrained environments, says the researcher. We are therefore taking up the principles of the encryption methods and protocols used today, such as HTTP, but taking into account the specific requirements of the IoT”.

The research team has developed a compression mechanism named SCHC (Static Context Header Compression, pronounced “chic”). It aims to improve the efficiency of encryption solutions and provide interoperability in low-power networks.

For this purpose, SCHC works on the headers of the usual Internet protocols (IP, UDP and CoAP), which contain various details: source address, destination address, location of the data to be read, etc. The particularity of this method is that it takes advantage of the specificity of the IoT: a simple connected object, such as a sensor, has far fewer functions than a smartphone. It is then possible to anticipate the type of data sent. “We can thus free ourselves from the redundancy of classic exchanges on the web, says Laurent Toutain. We then lose flexibility, which could be inconvenient for standard Internet use, but not for a sensor, which is limited in its applications”.

In this way, the team at IMT Atlantique has achieved significant results. It has managed to reduce the size of the headers traditionally sent, weighing 70-80 bytes, to only 2 bytes, and to 10 bytes in their encrypted version. “A quantity that is perfectly acceptable for a connected object and compatible with network architectures that consume very little energy,” concludes the researcher.

A protocol approved by the IETF

But what about that precious interoperability? With this objective, the authors of the study approached the IETF (Internet Engineering Task Force), the international organization for Internet standards. The collaboration has paid off, as SCHC has been approved by the organization and now serves as the global standard for compression. This recognition is essential, but is only a first step towards effective interoperability. How can we now make sure that manufacturers really integrate the protocol into their connected objects? For this, Laurent Toutain has partnered with Alexander Pelov, also a researcher at IMT Atlantic, in order to found the start-up company Acklio. The company works directly with industrialists and offers them solutions to integrate SCHC in their products. It thus intends to accelerate the democratization of the protocol, an effort supported in particular by  €2 million in funds raised at the end of 2019.

Read more on I’MTech Acklio: linking connected objects to the Internet

Nevertheless, industrialists remain to be convinced that the use of a standard is also in their interest. To this end, Acklio also aims to position SCHC among the protocols used within 5G. To achieve this, it will have to prove itself with the 3GPP (3rd Generation Partnership Project) which brings together the world’s leading telecommunications standards bodies. “A much more constraining process than that of the IETF,” however, warns Laurent Toutain.

Bastien Contreras

Eclairer boites noires, algorithms

Shedding some light on black box algorithms

/0 Comments/in , /by

In recent decades, algorithms have become increasingly complex, particularly through the introduction of deep learning architectures. This has gone hand in hand with increasing difficulty in explaining their internal functioning, which has become an important issue, both legally and socially. Winston Maxwell, legal researcher, and Florence d’Alché-Buc, researcher in machine learning, who both work for Télécom Paris, describe the current challenges involved in the explainability of algorithms.

What skills are required to tackle the problem of algorithm explainability?

Winston Maxwell: In order to know how to explain algorithms, we must draw on different disciplines. Our multi-disciplinary team, AI Operational Ethics, focuses not only on mathematical, statistical and computational aspects, but also on sociological, economic and legal aspects. For example, we are working on an explainability system for image recognition algorithms used, among other things, for facial recognition in airports. Our work therefore encompasses these different disciplines.

Why are algorithms often difficult to understand?

Florence d’Alché-Buc: Initially, artificial intelligence used mainly symbolic approaches, i.e., it simulated the logic of human reasoning. Logical rules, called expert systems, allowed artificial intelligence to make a decision by exploiting observed facts. This symbolic framework made AI more easily explainable. Since the early 1990s, AI has increasingly relied on statistical learning, such as decision trees or neural networks, as these structures allow for better performance, learning flexibility and robustness.

This type of learning is based on statistical regularities and it is the machine that establishes the rules which allow their exploitation. The human provides input functions and an expected output, and the rest is determined by the machine. A neural network is a composition of functions. Even if we can understand the functions that compose it, their accumulation quickly becomes complex. So a black box is then created, in which it is difficult to know what the machine is calculating.

How can artificial intelligence be made more explainable?

FAB: Current research focuses on two main approaches. There is explainability by design where, for any new constitution of an algorithm, explanatory output functions are implemented which make it possible to progressively describe the steps carried out by the neural network. However, this is costly and impacts the performance of the algorithm, which is why it is not yet very widespread. In general, and this is the other approach, when an existing algorithm needs to be explained, it is an a posteriori approach that is taken, i.e., after an AI has established its calculation functions, we will try to dissect the different stages of its reasoning. For this there are several methods, which generally seek to break the entire complex model down into a set of local models that are less complicated to deal with individually.

Why do algorithms need to be explained?

WM: There are two main reasons why the law stipulates that there is a need for the explainability of algorithms. Firstly, individuals have the right to understand and to challenge an algorithmic decision. Secondly, it must be guaranteed that a supervisory institution such as the  French Data Protection Authority (CNIL), or a court, can understand the operation of the algorithm, both as a whole and in a particular case, for example to make sure that there is no racial discrimination. There is therefore an individual aspect and an institutional aspect.

Does the format of the explanations need to be adapted to each case?

WM: The formats depend on the entity to which it needs to be explained: for example, some formats will be adapted to regulators such as the CNIL, others to experts and yet others to citizens. In 2015, an experimental service to deploy algorithms that detect possible terrorist activities in case of serious threats was introduced. For this to be properly regulated, an external control of the results must be easy to carry out, and therefore the algorithm must be sufficiently transparent and explainable.

Are there any particular difficulties in providing appropriate explanations?

WM: There are several things to bear in mind. For example, information fatigue: when the same explanation is provided systematically, humans will tend to ignore it. It is therefore important to use varying formats when presenting information. Studies have also shown that humans tend to follow a decision given by an algorithm without questioning it. This can be explained in particular by the fact that humans will consider from the outset that the algorithm is statistically wrong less often than themselves. This is what we call automation bias. This is why we want to provide explanations that allow the human agent to understand and take into consideration the context and the limits of algorithms. It is a real challenge to use algorithms to make humans more informed in their decisions, and not the other way around. Algorithms should be a decision aid, not a substitute for human beings.

What are the obstacles associated with the explainability of AI?

FAB: One aspect to be considered when we want to explain an algorithm is cyber security. We must be wary of the potential exploitation of explanations by hackers. There is therefore a triple balance to be found in the development of algorithms: performance, explainability and security.

Is this also an issue of industrial property protection?

WM: Yes, there is also the aspect of protecting business secrets: some developers may be reluctant to discuss their algorithms for fear of being copied. Another counterpart to this is the manipulation of scores: if individuals understand how a ranking algorithm, such as Google’s, works, then it would be possible for them to manipulate their position in the ranking. Manipulation is an important issue not only for search engines, but also for fraud or cyber-attack detection algorithms.

How do you think AI should evolve?

FAB: There are many issues associated with AI. In the coming decades, we will have to move away from the single objective of algorithm performance to multiple additional objectives such as explainability, but also equitability and reliability. All of these objectives will redefine machine learning. Algorithms have spread rapidly and have enormous effects on the evolution of society, but they are very rarely accompanied by instructions for their use. A set of adapted explanations must go hand in hand with their implementation in order to be able to control their place in society.

By Antonin Counillon

Also read on I’MTech: Restricting algorithms to limit their powers of discrimination

 

La Ruche à vélos, bicycle parking

La Ruche à Vélos is developing secure bicycle parking throughout France

/0 Comments/in , , , /by

Innovative and appropriate parking solutions must be created for the long-term development of cycling. The La Ruche à Vélos start-up incubated at IMT Atlantique offers an automated, secure and easy-to-use parking facility. This modular concept is connected to a mobile application and is intended for all users via acquisition by local authorities. For this solution, La Ruche à Vélos won the 2020 Bercy-IMT Innovation Award on February 2nd.

In 2020, many French people got back on their bikes. In its annual report published last October, the Vélo & Territoires association reported an average increase in bicycle use of 9% between January and September 2020 (compared to 2019) [1]. In a year strongly marked by strikes and the health crisis, exceptional circumstances strongly supported this trend. The attraction for bicycles shows no signs of slowing down. While local authorities support these practices, they also raise new issues in terms of security and parking. How many cyclists have already found their bike without a saddle, without a wheel, or perhaps not found their bike at all? To meet these challenges, the start-up La Ruche à Vélos, incubated at IMT Atlantique, proposes an innovative secure bicycle storage solution.

Automatic and secure parking

The increase in the number of cyclists is due in part to the emergence of electric bicycles. These bikes are heavier, bulkier and require a significant financial investment by their users. They therefore pose new constraints and require more security when parking. La Ruche à Vélos has developed a product that meets these expectations. Their solution consists of a secure bicycle parking facility which is connected to a mobile application. Its three founders were particularly attached to its ease of use. “It takes between 20 and 30 seconds to drop off or pick up a bike,” says Antoine Cochou, co-creator of the start-up. But how does it work?

The application allows the user to geolocate a parking facility with available spaces and to reserve one in advance. After identifying themselves on site, cyclists have access to a chamber, and deposit their bike on a platform before validating. There are also compartments available allowing users to recharge their batteries. Inside the parking facility, a machine stores the bike automatically. The facility covers several floors, thus saving ground space and facilitating integration of the system into the urban landscape. It can hold about 50 bikes over 24 square meters, dividing the bicycle parking space otherwise required on sidewalks by four! In addition, the size of the parking facility is flexible. The number of spaces therefore varies according to the order.

In June 2021, a first prototype of about ten spaces will be installed in the city of Angers. The young innovators hope to collect enough feedback from users to improve their next product. Two more facilities are planned for the year. They will have 62 to 64 spaces. “Depending on the location, a balance must be struck between user waiting time and the demand for services. These two parameters are related to the number of sites and the flow of users at peak times (train station, shops, etc.),” says Antoine Cochou.

Strategic locations with adapted subscriptions

La Ruche à Vélos is aimed directly at local authorities who can integrate this solution into their mobility program. It also targets businesses and real estate developers wishing to offer an additional service to their employees or future residents. Depending on the needs, the parking facilities could therefore be installed in different strategic locations. “Local authorities are currently focusing on railway stations and city centers, but office or residential areas are also being considered,” says Antoine Cochou. Each zone has its own target and therefore its own form of subscription. In other words, one-off parking in the city, daytime offers for offices, and evening and weekend passes for residents.

Initially, subscriptions for the prototype installed in Angers will be managed by the start-up. However, future models are expected to couple parking passes with local public transit passes. Subscriptions will thus be taken care of by the cities. The start-up will focus on maintenance support. “In this sense, our next models will be equipped with cameras and it will be possible to control them remotely,” says Maël Beyssat, co-creator of La Ruche à Vélos. Communities will have a web interface to monitor the condition and operating status of the parking facility (rate of use, breakdowns, availability, etc.)

For the future, the company is considering the installation of solar panels to offer a zero-electricity consumption solution. Finally, other locations could be considered outside of popular touring sites on cycle routes.

[1] Result obtained with the help of sensors measuring the number of bikes going past.

By Anaïs Culot

David Gesbert, PERFUME

PERFUME: a scent of cooperation for the networks of the future

/0 Comments/in , /by

The ERC PERFUME project, led by EURECOM researcher David Gesbert and ending in 2020, resulted in the development of algorithms for local decision making in the mobile network. This research was tested on autonomous drones, and is particularly relevant to the need for connected robotics in the post-5G world.

Now that 5G is here, who’s thinking about what comes next? The team working with David Gesbert, a researcher specializing in wireless communication systems at EURECOM, has just completed its ERC PERFUME project on this subject. So what will wireless networks look like by 2030? While 5G is based on the centralization of calculations in the cloud, the networks of the future will require, on the contrary, a distributed network. By this, we mean the emergence of a more cooperative network. “In the future, the widespread use of robotic objects and devices to perform autonomous tasks will increase the need for local decision making, which is difficult in a centralized system,” says Gesbert. Nevertheless, the objective remains the same: optimizing the quality of the network. This is especially important since the increase in connected devices may cause more interference and therefore affect the quality of the information exchanged.

Why decentralize decision making on the network?

Under 5G, every device that is connected to the network can send measurements to the cloud. The cloud has a very high computing capacity, enabling it to process an immeasurable amount of data, before sending instructions back to devices (a tablet, cell phone, drone, etc.). However, these information transfers take time, which is a very valuable commodity for connected robotics applications or critical missions. Autonomous vehicles, for example, must make instant decisions in critical situations. “In the context of real-time applications, the response speed of the network must be optimized. Decentralizing decisions closer to the base stations is precisely the solution that was studied in our PERFUME project,” explains David Gesbert. As 5G is not yet equipped to meet this constraint, we have to introduce new evolutions of the standard.

EURECOM’s researchers are thus relying on cooperation and coordination of the computing capabilities of local terminals such as our cell phones. By exchanging information, these terminals could coordinate in the choice of their power and transmission frequency, which would limit the interference that would limit the flow rates, for example. They would no longer focus solely on their local operations, but would participate in the overall improvement of the quality of the network. A team effort that would manifest itself at the user level by sending files faster or providing better image quality during a video call. However, although possible, this collaboration remains difficult to implement.

Towards more cooperative wireless networks

Distributed networks pose a major problem: access to information from one device to another is incomplete. “Our problem of exchanging information locally can be compared to a soccer team playing blindfolded. Each player only has access to a noisy piece of information and doesn’t know where the other team members are in their attempt to score the goal together”, says David Gesbert. Researchers then develop so-called robust decision-making algorithms. Their objective? To allow a set of connected devices to process this noisy information locally. “Networks have become too complicated to be optimized by conventional mathematical solutions, and they are teeming with data. This is why we have designed algorithms based on signal processing but also on machine learning,” continues the researcher.

These tools were then tested in a concrete 5G network context in partnership with Ericsson. “The objective was for 5G cells to coordinate on the choice of directional beams of MIMO (multi-input multi-output) antennas to reduce interference between them,” says the researcher. These smart antennas, deployed as part of 5G, are increasingly being installed on connected devices. They perform “beamforming”, which means that they direct a radio signal in a specific direction – rather than in all directions – thus improving the efficiency of the signal. These promising results have opened the door to large-scale tests on connected robotics applications, the other major focus of the ERC project. EURECOM has thus experimented with certain algorithms on autonomous drones.

Drones at the service of the network?

Following a disaster such as an avalanche, a tsunami or an earthquake, part of the ground network infrastructure may be destroyed and access to the network may be cut off. It would then be possible to replicate a temporary network architecture on site using a fleet of drones serving as air relays. On the EURECOM campus, David Gesbert’s team has developed prototypes of autonomous drones connected to 5G. These determine a strategic flight position and their respective positions in order to solve network access problems for users on the ground. The drones then move freely and recalculate their optimal placement according to the user’s position.  This research notably received the prize for the best 2019 research project, awarded by the Provence-Alpes-Côte d’Azur region’s Secure Communicating Solutions cluster.

This solution could be considered in the context of rescue missions and geolocalization of missing persons. However, several challenges need to be addressed for this method to develop. Indeed, current regulations prohibit the theft of autonomous aircraft. In addition, they have a flight time of about 30 minutes, which is still too short to offer sustainable solutions.

“This research is also adapted to issues relating to autonomous cars, adds David Gesbert. For example, when two vehicles arrive at an intersection, a protocol for coordination must be put in place to ensure that the vehicles cross the intersection as quickly as possible and with the lowest likelihood of collision.” In addition, medicine and connected factories would also be targets for application of distributed networks. As for the integration of this type of standard in the future 6G, it will depend on the interests of industrial players in the years to come.

By Anaïs Culot

Learn more about the ERC PERFUME project

RAMP UP Seed

Supporting companies in the midst of crisis

/0 Comments/in , , /by

The RAMP-UP Seed project is one of 9 new projects to have been selected by the German-French Academy for the Industry of the Future as part of the “Resilience and Sustainability for the Industry of the Future” call for projects. It focuses on helping companies adapt their production capacities to respond to crisis situations. The project relies on two main areas of expertise to address this issue: ramp-up management and artificial intelligence (AI). Khaled Medini and Olivier Boissier, researchers at Mines Saint-Étienne,[1] a partner of the project, present the issues.

Can you describe the context for the RAMP-UP Seed project?

Khaled Medini The RAMP-UP Seed project is one of 9 new projects to have been selected by the German-French Academy for the Industry of the Future (GFA) for the call for projects on the sustainability and resiliency of companies in the industry of the future. This project is jointly conducted by Mines Saint-Étienne and TUM (Technische Universität München), and is a continuation of work carried out on diversity management, ramp-up management, and multi-agent systems at Institut Fayol related to the industry of the future.

What is the project’s goal?

KM The health crisis has highlighted the limitations of current industrial models when it comes to providing a quick response to market demands in terms of quality and quantity, and production constraints related to crisis situations. Ramp-up and ramp-down management is a key to meeting these challenges. The goal of RAMP-UP Seed is to establish a road map for developing a tool-based methodology in order to increase companies’ sustainability and resilience specifically by targeting the adaptation phase and production facilities.

How do you plan to achieve this? What are the scientific obstacles you must overcome?

Olivier Boissier The project addresses issues related to the topics of production systems and artificial intelligence. The goal is to remedy a lack of methodology guides and tools for strengthening companies’ sustainability and resilience. Two main actions will be prioritized for this purpose during the initial seed stage:

  • An analysis of existing approaches and identification with industrial stakeholders of needs and use cases, which will be conducted jointly with two partners;
  • Development of a proposal for a collaborative project involving Franco-German academic and industrial partners in order to respond to European calls for projects.

From an operational standpoint, work meetings and workshops are held regularly with teams from Mines Saint-Étienne and the TUM in a spirit of collaboration.

Who are the partners involved in this project and what are their respective roles?

KM We started RAMP-UP Seed in partnership with the TUM Institute of Automation and Information Systems with a focus on two main areas: ramp-up management and artificial intelligence. Expertise from the Territoire and IT’M Factory platforms from Institut Henri Fayol, and TUM platforms will be used to develop this dynamic further.

Who will benefit from the methods and tools developed by RAMP-UP Seed?

OB The purpose of the multi-agent optimization and simulation tools and industrial management tools to be developed through this project is to provide decision-making tools for exploring, testing and better managing ramp-up in the manufacturing and service sectors. Special attention will also be given to the health crisis, with a focus on the health sector.

What are the next big steps for the project?

KM RAMP-UP Seed is a seed project. In addition to analyzing current trends, one of the key goals is to develop joint responses to calls for projects in the fields of artificial intelligence and industrial management.

[1] Khaled Medini is a researcher at the Laboratory of Informatics, Modeling and Optimization of Systems (LIMOS), joint research unit UCA/Mines Saint-Étienne/CNRS 6158). Olivier Boissier is a researcher at Hubert Curien Laboratory, joint research unit CNRS 5516/Mines Saint-Étienne).

Interview by Véronique Charlet

Facial recognition: what legal protection exists?

/0 Comments/in , , /by

Over the past decade, the use of facial recognition has developed rapidly for both security and convenience purposes. This biometrics-based technology is used for everything from video surveillance to border controls and unlocking digital devices. This type of data is highly sensitive and is subject to specific legal framework. Claire Levallois-Barth, a legal researcher at Télécom Paris and coordinator of the Values and Policies of Personal Information Chair at IMT provides the context for protecting this data.

What laws govern the use of biometric data?

Claire Levallois-Barth: Biometric data “for the purpose of uniquely identifying a natural person” is part of a specific category defined by two texts adopted by the 27 Member States of the European Union in April 2016, the General Regulation Data Protection Regulation (GDPR) and the Directive for Police and Criminal Justice Authorities. This category of data is considered highly sensitive.

The GDPR applies to all processing of personal data in both private and public sectors.

The Directive for Police and Criminal Justice Authorities pertains to processing carried out for purposes of prevention, detection, investigation, and prosecution of criminal offences or the execution of criminal penalties by competent authorities (judicial authorities, police, other law enforcement authorities). It specifies that biometric data must only be used in cases of absolute necessity and must be subject to provision of appropriate guarantees for the rights and freedoms of the data subject. This type of processing may only be carried out in three cases: when authorized by Union law or Member State law, when related to data manifestly made public by the data subject, or to protect the vital interests of the data subject or another person.

What principles has the GDPR established?

CLB: The basic principle is that collecting and processing biometric data is prohibited due to significant risks of violating basic rights and freedoms, including the freedom to come and go anonymously. There are, however, a series of exceptions. The processing must fall under one of these exceptions (express consent from the data subject, protection of his or her vital interests, conducted for reasons of substantial public interest) and respect all of the obligations established by the GDPR. The key principle is that the use of biometric data must be strictly necessary and proportionate to the objective pursued. In certain cases, it is therefore necessary to obtain the individual’s consent, even when the facial recognition system is being used on an experimental basis. There is also the minimization principle, which systematically asks, “is there any less intrusive way of achieving the same goal?” In any case, organizations must carry out an impact assessment on people’s rights and freedoms.

What do the principles of proportionality and minimization look like in practice?

CLB: One example is when the Provence-Alpes-Côte d’Azur region wanted to experiment with facial recognition at two high schools in Nice and Marseille. The CNIL ruled that the system involving students, most of whom were minors, for the sole purpose of streamlining and securing access, was not proportionate to these purposes. Hiring more guards or implementing a badge system would offer a sufficient solution in this case.

Which uses of facial recognition have the greatest legal constraints?

CLB: Facial recognition can be used for various purposes. The purpose of authentication is to verify whether someone is who he or she claims to be. It is implemented in technology for airport security and used to unlock your smartphone. These types of applications do not pose many legal problems. The user is generally aware of the data processing that occurs, and the data is usually processed locally, by a card for example.

On the other hand, identification—which is used to identify one person within a group—requires the creation of a database that catalogs individuals. The size of this database depends on the specific purposes. However, there is a general tendency towards increasing the number of individuals. For example, identification can be used to find wanted or missing persons, or to recognize friends on a social network. It requires increased vigilance due to the danger of becoming extremely intrusive.

Facial recognition has finally provided a means of individualizing a person. There is no need to identify the individual–the goal is “simply” to follow people’s movements through the store to assess their customer journey or analyze their emotions in response to an advertisement or while waiting at the checkout. The main argument advertisers use to justify this practice is that the data is quickly anonymized, and no record is kept of the person’s face. Here, as in the case of identification, facial recognition usually occurs without the person’s knowledge.

How can we make sure that data is also protected internationally?

CLB: The GDPR applies in the 27 Member States of the European Union which have agreed on common rules. Data can, however, be collected by non-European companies. This is the case for photos of European citizens collected from social networks and news sites. This is one of the typical activities of the company Clearview AI, which has already established a private database of 3 billion photos.

The GDPR lays down a specific rule for personal data leaving European Union territory: it may only be exported to a country ensuring a level of protection deemed comparable to that of the European Union. Yet few countries meet this condition. A first option is therefore for the data importer and exporter to enter into a contract or adopt binding corporate rules. The other option, for data stored on servers on U.S. territory, was to build on the Privacy Shield agreement concluded between the Federal Trade Commission (FTC) and the European Commission. However, this agreement was invalidated by the Court of Justice of the European Union in the summer of 2020. We are currently in the midst of a legal and political battle. And the battle is complicated since data becomes much more difficult to control once it is exported. This explains why certain stakeholders, such as Thierry Breton (the current European Commissioner for Internal Market), have emphasized the importance of fighting to ensure European data is stored and processed in Europe, on Europe’s own terms.

Despite the risks and ethical issues involved, is facial recognition sometimes seen as a solution for security problems?

CLB: It can in fact be a great help when implemented in a way that respects our fundamental values. It depends on the specific terms. For example, if law enforcement officers know that a protest will be held, potentially involving armed individuals, at a specific time and place, facial recognition can prove very useful at that specific time and place. However, it is a completely different scenario if it is used constantly for an entire region and entire population in order to prevent shoplifting.

This summer, the London Court of Appeal ruled that an automatic facial recognition system used by Welsh police was unlawful. The ruling emphasized a lack of clear guidance on who could be monitored and accused law enforcement officers of failing to sufficiently verify whether the software used had any racist or sexist bias. Technological solutionism, a school of thought emphasizing new technology’s capacity to solve the world’s major problems, has its limitations.

Is there a real risk of this technology being misused in our society?

CLB: A key question we should ask is whether there is a gradual shift underway, caused by an accumulation of technology deployed at every turn. We know that video-surveillance cameras are installed in public roads, yet we do not know about additional features that are gradually added, such as facial recognition or behavioral recognition.  The European Convention of Human Rights, GDPR, the Directive for Police and Criminal Justice Authorities, and the CNIL provide safeguards in this area.

However, they provide a legal response to an essentially political problem. We must prevent the accumulation of several types of intrusive technologies that come without prior reflection on the overall result, without taking a step back to consider the consequences. What kind of society do we want to build together? Especially within the context of a health and economic crisis. The debate on our society remains open, as do the means of implementation.

Interview by Antonin Counillon

Hack drone attaque UAVs

Hacked in mid-flight: detecting attacks on UAVs

/0 Comments/in , /by

A UAV (or drone) in flight can fall victim to different types of attacks. At Télécom SudParis, Alexandre Vervisch-Picois is working on a method for detecting attacks that spoof drones concerning their position. This research could be used for both military and civilian applications.

He set out to deliver your package one morning, but it never arrived. Don’t worry, nothing has happened to your mailman. This is a story about an autonomous drone. These small flying vehicles are capable of following a flight path without a pilot, and are now ahead of the competition in the race for the fastest delivery.

While drone deliveries are technically possible, for now they remain the stuff of science fiction in France. This is due to both legal reasons and certain vulnerabilities in these systems. At Télécom SudParis, Alexandre Vervisch-Picois, a researcher specialized in global navigation satellite systems (GNSS), and his team are working with Thales to detect what is referred to as “spoofing” attacks. In order to prevent these attacks, researchers are studying how they work, with the goal of establishing protocol to help detect them.

How do you spoof a drone?

In order to move around independently, a drone must know its position and the direction in which it is moving. It therefore receives continuous signals from a satellite constellation which enables it to calculate the coordinates of its position. These can then be used to follow a predefined flight path by moving through a succession of waypoints until it reaches its destination. However, the drone’s heavy reliance on satellite geolocation to find its way makes it vulnerable to cyber attacks. “If we can succeed in getting the drone to believe it is somewhere other than its actual position, then we can indirectly control its flight path,” Alexandre Vervisch-Picois explains. This flaw is all the more critical given that the drones’ GPS receivers can be easily deceived by false signals transmitted at the same frequency as those of the satellites.

This is what the researchers call a spoofing attack. This type of cyber attack is not new. It was used in 2011 by the Iranian army to capture an American stealth drone that flew over its border. The technique involves transmitting a sufficiently powerful false radio frequency to replace the satellite signal picked up by the drone. This spoofing technique doesn’t cancel the drone’s geolocation capacities as a scrambler would. Instead, it forces the GPS receiver to calculate an incorrect position, causing it to deviate from its flight path. “For example, an attacker who succeeds in identifying the next waypoint can then determine a wrong position to be sent in order to lead the drone right to a location where it can be captured,” the researcher explains.

Resetting the clocks

Several techniques can be used to identify these attacks, but they often require additional costs, both in terms of hardware and energy.Through the DIGUE project (French acronym for GNSS Interference Detection for Autonomous UAV)[1] conducted with Thales Six, Alexandre Vervisch-Picois and his team have developed a method for detecting spoofing attempts. “Our approach uses the GPS receivers present in the drones, which makes this solution less expensive,” says the researcher. This is referred to as the “clock bias” method. Time is a key parameter in satellite position calculations. The satellites have their time base and so does the GPS receiver. Therefore, once the GPS receiver has calculated its position, it measures the “bias”, which is the difference between these two time bases.  However, when a spoofing attack occurs, the researchers observed variations in this calculation in the form of a jump. The underlying reason for this jump is that the spoofer has its own time base, which is different from that of the satellites. “In practice, it is impossible for the spoofer to use the same clock as a satellite. All it can do is move closer to the time base, but we always notice a jump,” Alexandre Vervisch-Picois explains. To put it simply, satellites and spoofer are not set to the same time.

One advantage of this method is that it does not require any additional components or computing power to retrieve the data, since they are already present in the drone. It also does not require expensive signal processing analyses in order to study the information received by the drone–which is another defense method used to determine whether or not a signal originated from a satellite.

But couldn’t the attacker work around this problem by synchronizing with the satellites’ time setting? “It is very rare but still possible in the case of a very sophisticated spoofer. This is a classic example of measures and countermeasures, exemplified in interactions between a sword and a shield. In response to an attack, we set up defense systems and the threats become more sophisticated to bypass them,” the researcher explains. This is one reason why research in this area has so much to offer.

After obtaining successful results in the laboratory, the researchers are now planning to develop an algorithm based on time bias monitoring. This could be implemented on a flying drone for a test with real conditions.

What happens after an attack is detected?

Once the attack has been detected, the researchers try to locate the source of the false signal in order to find the attacker. To do so, they propose using a fleet of connected drones. The idea is to program movements within the fleet in order to determine the angle of arrival for the false signal. One of the drones would then send a message to the relevant authorities in order to stop the spoofer. This method is still in its infancy and is expected to be further developed with Thales in a military context with battlefield situations in which the spoofer must be eliminated. But in the context of a parcel delivery, what could be used to defend a single drone? “There could be a protocol involving rising to a higher altitude to move out of the spoofer’s range, which can reach up to several kilometers. But it would certainly not be as easy to escape its influence,” the researcher says. Another alternative could be to use signal processing methods, but these solutions would increase the costs associated with the device. “If too much of the drone’s energy is required for its protection, we need to ask whether this mode of transport is helpful and perhaps consider other more conventional methods, which are less burdensome to implement,” says Alexandre Vervisch-Picois.

[1] Victor Truong’s thesis research

Anaïs Culot

The Alicem app: a controversial digital authentication system

/0 Comments/in , , /by

Laura Draetta, Télécom Paris – Institut Mines-Télécom and Valérie Fernandez, Télécom Paris – Institut Mines-Télécom

[dropcap]S[/dropcap]ome digital innovations, although considered to be of general interest, are met with distrust. A responsible innovation approach could anticipate and prevent such confidence issues.

“Alicem” is a case in point. Alicem is a smartphone app developed by the State to offer the French people a national identity solution for online administrative procedures. It uses face recognition as a technological solution to activate a user account and allow the person to prove their digital identity in a secure way.

After its authorization by decree of May 13, 2019 and the launch of the experimentation of a prototype among a group of selected users a few months later, Alicem was due to be released for the general public by the end of 2019.

However, in July of the same year, La Quadrature du Net, an association for the defense of rights and freedoms on the Internet, filed an appeal before the Council of State to have the decree authorizing the system annulled. In October 2019, the information was relayed in the general press and the app was brought to the attention of the general public. Since then, Alicem has been at the center of a public controversy surrounding its technological qualities, potential misuses and regulation, leading to it being put on hold to dispel the uncertainties.

At the start of the summer of 2020, the State announced the release of Alicem for the end of the autumn, more than a year later than planned in the initial roadmap. Citing the controversy on the use of facial recognition in the app, certain media actors argued that it was still not ready: it was undergoing further ergonomic and IT security improvements and a call to tender was to be launched to build “a more universal and inclusive offer” incorporating, among other things, alternative activation mechanisms to facial recognition.

Controversy as a form of “informal” technology assessment

The case of Alicem is similar to that of other controversial technological innovations pushed by the State such as the Linky meters, 5G and the StopCovid app, and leads us to consider controversy as a form of informal technology assessment that defies the formal techno-scientific assessments that public decisions are based on. This also raises the issue of a responsible innovation approach.

Several methods have been developed to evaluate technological innovations and their potential effects. In France, the Technology Assessment – a form of political research that examines the short- and long-term consequences of innovation – is commonly used by public actors when it comes to technological decisions.

In this assessment method, the evaluation is entrusted to scientific experts and disseminated among the general public at the launch of the technology. The biggest challenge with this method is supporting the development of public policies while managing the uncertainties associated with any technological innovation through evidence-based rationality. It must also “educate” the public, whose mistrust of certain innovations may be linked to a lack of information.

The approach is perfectly viable for informing decision-making when there is no controversy or little mobilization of opponents. It is less pertinent, however, when the technology is controversial. A technological assessment focused exclusively on scholarly expertise runs the risk of failing to take account of all the social, ethical and political concerns surrounding the innovation, and thus not being able to “rationalize” the public debate.

Participation as a pillar of responsible innovation

Citizen participation in technology assessment – whether to generate knowledge, express opinions or contribute to the creation and governance of a project – is a key element of responsible innovation.

Participation may be seen as a strategic tool for “taming” opponents or skeptics by getting them on board or as a technical democracy tool that gives voice to ordinary citizens in expert debates, but it is more fundamentally a means of identifying social needs and challenges upstream in order to proactively take them into account in the development phase of innovations.

In all cases, it relies on work carried out beforehand to identify the relevant audiences (users, consumers, affected citizens etc.) and choose their spokespersons. The definition of the problem, and therefore the framework of the response, depends on this identification. The case of Linky meters is an emblematic example: anti-radiation associations were not included in the discussions prior to deployment because they were not deemed legitimate to represent consumers; consequently, the figure of the “affected citizen” was nowhere to be seen during the discussions on institutional validation but is now at the center of the controversy.

Experimentation in the field to define problems more effectively

Responsible innovation can also be characterized by a culture of experimentation. During experimentation in the field, innovations are confronted with a variety of users and undesired effects are revealed for the first time.

However, the question of experimentation is too often limited to testing technical aspects. In a responsible innovation approach, experimentation is the place where different frameworks are defined, through questions from users and non-users, and where tensions between technical efficiency and social legitimacy emerge.

If we consider the Alicem case through the prism of this paradigm, we are reminded that technological innovation processes carried out in a confined manner – first of all through the creation of devices within the ecosystem of paying clients and designers and then through the experimentation of the use of artifacts already considered stable – inevitably lead to acceptability problems. Launching a technological innovation without participation in its development by the users undoubtedly makes the process faster, but may cost its legitimization and even lead to a loss of confidence for its promoters.

In the case of Alicem, the experiments carried out among “friends and family”, with the aim of optimizing the user experience, could be a case in point. This experimentation was focused more on improving the technical qualities of the app than on taking account of its socio-political dimensions (risk of infringing upon individual freedoms and loss of anonymity etc.). As a result, when the matter was reported in the media it was presented through an amalgamation of face recognition technology use cases and anxiety-provoking arguments (“surveillance”, “freedom-killing technology”, “China”, “Social credit” etc.). Without, however, presenting the reality of more common uses of facial recognition which carry the same risks as those being questioned.

These problems of acceptability encountered by Alicem are not circumstantial ones unique to a specific technological innovation, but must be understood as structural markers of the contemporary social functioning. For, although the “unacceptability” of this emerging technology is a threat for its promoters and a hindrance to its adoption and diffusion, it is above all indicative of a lack of confidence in the State that supersedes the reality of the intrinsic qualities of the innovation itself.

This text presents the opinions stated by the researchers Laura Draetta and Valérie Fernandez during their presentation at the Information Mission on Digital Identity of the National Assembly in December 2019. It is based on the case of the biometric authentication app Alicem, which sparked controversy in the public media sphere from the first experiments.

Laura Draetta, a Lecturer in Sociology, joint holder of the Responsibility for Digital Identity Chair, Research Fellow Center for Science, Technology, Medicine & Society, University of California, Berkeley, Télécom Paris – Institut Mines-Télécom and Valérie Fernandez, Professor of Economics, Holder of the Responsibility for Digital Identity chair, Télécom Paris – Institut Mines-Télécom

This article was republished from The Conversation under the Creative Commons license. Read the original article here.

 

Étienne Perret, IMT-Académie des sciences Young Scientist prize

/0 Comments/in , , /by

What if barcodes disappeared from our supermarket items? Étienne Perret, a researcher in radio-frequency electronics at Grenoble INP, works on identification technologies. His work over recent years has focused on the development of RFID without electronic components, commonly known as chipless RFID. The technology aims to offer some of the advantages of classical RFID but at a similar cost to barcodes, which are more commonly used in the identification of objects. This research is very promising for use in product traceability and has earned Étienne Perret the 2020 IMT-Académie des sciences Young Scientist Prize.

Your work focuses on identification technologies: what is it exactly?

Étienne Perret: The identification technology most commonly known to the general public is the barcode. It is on every item we buy. When we go to the checkout, we know that the barcode is used to identify objects. Studies estimate that 70% of products manufactured across the world have a barcode, making it the most widely used identification technique. However, it is not the only one, there are other technologies such as RFID (radio frequency identification). It is what is used on contactless bus tickets, ski passes, entry badges for certain buildings, etc. It is a little more mysterious, it’s harder to see what’s behind it all. That said, the idea behind it is the same, regardless of the technology. The aim is to identify an item at short or medium range.

What are the current challenges surrounding these identification technologies?

EP: In lots of big companies, especially Amazon, object traceability is essential. They often need to be able to track a product from the different stages of manufacturing right through to its recycling. Each product therefore has to be able to be identified quickly. However, both of the current technologies I mentioned have limitations as well as advantages. Barcodes are inexpensive, can be printed easily but store very little information and often require human input between the scanner and the code to make sure it is read correctly. What is more, barcodes have to be visible in order to be read, which has an effect on the integrity of the product to be traced.

RFID, on the other hand, uses radio waves that pass through the material, allowing us to identify an object already packaged in a box from several meters away. However, this technology is costly. Although an RFID label only costs a few cents, it is much more expensive than a barcode. For a company that has to label millions of products a year, the difference is huge, in particular when it comes to labeling products that are worth no more than a few cents themselves.

What is the goal of your research in this context?

EP: My aim is to propose a solution in between these two technologies. At the heart of an RFID tag there is a chip that stores information, like a microprocessor. The idea I’m pursuing with my colleagues at Grenoble INP is to get rid of this chip, for economic and environmental reasons. The other advantage that we want to keep is the barcode’s ease of printing. To do so, we base our work on an unusual approach combining conductive ink and geometric labels.

How does this approach work?  

EP: The idea is that each label has a unique geometric form printed in conductive ink. Its shape means that the label reflects radio frequency waves in a unique way. After that, it is a bit like a radar approach: a transmitter emits a wave, which is reflected by its environment, and the label returns the signal with a unique signature indicating its presence. Thanks to a post-processing stage, we can then recover this signature containing the information on the object.

Why is this chipless RFID technology so promising?

EP: Economically speaking, the solution would be much more advantageous than an RFID chip and could even rival the cost of a barcode. Compared to the latter, however, there are two major advantages. First of all, this technology can read through materials, like RFID. Secondly, it requires a simpler process to read the label. When you go through the supermarket checkout, the product has to be at a certain angle so that the code is facing the laser scanner. That is another problem with barcodes: a human operator is often required to carry out the identification and while it is possible to do without, it requires very expensive automated systems. Chipless RFID technology is not perfect, however, and certain limitations must be accepted, such as the reading distance, which is not the same as for conventional RFID which can reach several meters using ultra high frequency waves.

One of the other advantages of RFID is the ability to reprogram it: the information contained in an RFID tag can be changed. Is this possible with the chipless RFID technology you are developing?

EP: That is indeed one of the current research projects. In the framework of the ERC ScattererID project, we are seeking to develop the concept of rewritable chipless labels. The difficulty is obviously that we can’t use electronic components in the label. Instead, we’re basing our approach on CBRAM (conductive-bridging RAM) which is used for specific types of memories. It works by stacking three layers: metal-dielectric material-metal. Imagine a label printed locally with this type of stack. By applying a voltage to the printed pattern we can modify its properties and thus change the information contained in the label.

Does this research on chipless RFID technology have other applications than product traceability and identification?

EP: Another line of research we are looking into is using these chipless labels as sensors. We have shown that we can collect and report information on physical quantities such as temperature and humidity. For temperature, the principle is based on the ability to measure the thermal expansion of the materials that make up the label. The material “expands” by a few tens of microns. The label’s radiofrequency signature changes, and we are able detect these very subtle variations. In another field, this level of precision, obtained using radio waves, which are wireless, allows the label to be located and its movements detected. Based on this principle, we are currently also studying gestural recognition to allow us to communicate with the reader through the label’s movements.

The transfer of this technology to industry seems inevitable: where do you stand on this point?

EP: A recent project with an industrial actor led to the creation of the start-up Idyllic Technology, which aims to market chipless RFID technology to industrial firms. We expect to start presenting our innovations to companies during the course of next year. At present, it is still difficult for us to say where this technology will be used. There’s a whole economic dimension which comes into play, which will be decisive in its adoption. What I can say, though, is that I could easily see this solution being used in places where the barcode isn’t used due to its limitations, but where RFID is too expensive. There’s a place between the two, but it’s still too early to say exactly where.