Code injection remains one of the top vulnerabilities in computer
programs, but conventional mitigations (static analysis, fuzzing,
machine learning) rely mostly on known software flaws or empirical
analysis. A great deal of possible injections remains undetected and
prime targets for attackers.
A team of researchers have developed a formal verification
approach that can check and list all unsafe functions/methods in a
source code.
Tor is the most famous network available for anonymous browsing.
But it suffers many limitations that prevents it being widely used :
latency, ping and lack of resiliency. It also not adapted for one-tomany communications.
QasTor allows anonymous communication with a server/a website.
It uses ABE, a flexible encryption system that allows one-to-many
communication, marking it particularly applicable to streaming
applications (unlike Tor from which it is inspired).
Voice as a human-machine interface is becoming ubiquitous. But
traditional speech-to-text analysis fail to recognize contextual
meaning (emotions or speech intentions), that are precious to
improve user experience.
VoToNa is an audio signal processing mobile toolkit recognizing
prosody features of the voice.
SSCCP is an automated software tool to generate very large sets of
yearly capacity planning scenarios, providing help to identify critical
weaknesses in the supply chain.
DeepEdge uses deep learning for detecting edges in a 3D point
cloud. From a dataset acquired by photogrammetry, LIDAR or depth
camera, it identifies all points belonging to an edge or close to it.
COBANN is a FPGA or SoC coprocessor architecture on which any
type of neural network can be configured, trained and updated onthe-fly. No new implementation is required.
An unsupervised & bio-inspired machine learning algorithm for movement recognition pre-processing in event-based vision.
Our solution: STDP Motion is a Spiking Neural Network (SNN) dedicated to event-based vision. It can be used to predict the trajectory and landing position of a moving object, or any recuring event given an initial information.
La preuve de son identité doit être faite régulièrement en ligne tout en protégeant l’ensemble des acteurs de risques grandissants (fraude, usurpation d’identité, etc.)
Notre solution : VCUCIM sécurise et fluidifie ces échanges en rendant à l’utilisateur la maitrise des données qu’il transmet aux différentes organisations et fournisseurs de service sur internet.
Plus de 50% des masses d’eau de surface et près de 40% des masses d’eau souterraine en France n’atteignent pas l’objectif de la Directive Cadre sur l’Eau (DCE), les nitrates et pesticides étant les paramètres les plus déclassants. La régulation naturelle provenant des écosystèmes est un vecteur important de régénération mais il reste difficile pour les gestionnaires de caractériser et quantifier son impact sur le long terme.
Notre solution offre un système de visualisation de données via une plateforme web associée a un SIG (Système d’Information Géographique).
Predictive maintenance relies on clustering algorithms to track unwanted changes in industrial processes and infrastructures. However, most of them tend to fail to detect slow deviations indicating a gradual deterioration of the system.
Our solution: Dyclee, a clustering algorithm developed at LAAS-CNRS offers remarkable capabilities due to its ability to dynamically track smooth and/or abrupt changes in evolving conditions.
