Detection | EPFL Graph Search

In general, detection is the action of accessing information without specific cooperation from with the sender. In the history of radio communications, the term "detector" was first used for a device that detected the simple presence or absence of a radio signal, since all communications were in Morse code. The term is still in use today to describe a component that extracts a particular signal from all of the electromagnetic waves present. Detection is usually based on the frequency of the carrier signal, as in the familiar frequencies of radio broadcasting, but it may also involve filtering a faint signal from noise, as in radio astronomy, or reconstructing a hidden signal, as in steganography. In optoelectronics, "detection" means converting a received optical input to an electrical output. For example, the light signal received through an optical fiber is converted to an electrical signal in a detector such as a photodiode. In steganography, attempts to detect hidden signals

Evanescent wave gas sensing based on modulation spectroscopy

Daniel Salzmann

In recent years, optical gas detectors based on a tunable diode laser absorption spectrometer has become more and more important for monitoring trace gases in atmospheric, medical, and industrial processes. In the medical field, blood gases, essentially pO2 and pCO2, are vital parameters for patient respiratory status monitoring. Typically, hypoxaemia or hypercapnia can be very harmful, especially for neonates, and requires continuous monitoring for rapid clinical intervention. Transcutaneous blood gases, permeating through the skin, can be continuously measured in patients by electrochemical sensors applied to the surface of the skin. These sensors suffer specific instability and must be re-calibrated. For this reason, the present thesis explores a new concept of gas sensors, applying light absorption at the surface of an integrated optics waveguide and using the advantages of a modulated spectroscopy technique for sensitivity and selectivity enhancement. In other words, the present thesis sets a mathematical model and demonstrates experimentally a new concept of evanescent wave gas sensing based on modulation spectroscopy. The field of application considered in this book is restricted to gas detection, however the concept can also be applied to the detection of liquids. In particular, we demonstrate the potential applicability of the concept to transcutaneous blood gas monitoring, and especially carbon dioxide monitoring. The system transfer function modelization of the concept includes carbon dioxide absorption spectrum in the near-infrared region (1.6µm), laser diode modulation spectroscopy assuming optical frequency modulation (FM) as well as intensity modulation (IM) induced by current modulation, and integrated optics sensor-head transmission. The modelization evidences that the silicon nitride technology has the advantage of: high evanescent wave sensitivity (8% for the selected TE mode), low propagation loss, possibility of strong curvatures for miniaturization and availability of fiberto- waveguide mode matching technique (taper). In the modelization, we evidence that the optimal waveguide length equals the inverse of the propagation loss which is mainly limited by the waveguide scattering. Also, we demonstrate that the implementation of a fiber-to-waveguide mode matching taper at both waveguide ends, drastically decreases the resonance limitation of the system resolution. Experimentally, the transfer function of the concept is validated and characterized based on a free-space modulation spectroscopy setup whose performance, in terms of second harmonic resolution, is first measured at 3.55 · 10-6 1/√Hz corresponding to 0.275 mmHg/√Hz CO2 partial pressure. In the chosen region of 200-400MHz of frequency modulation, we confirm that the 1/f-noise is negligible and the dominating amplifier noise can be approximated by the thermal noise. In the case of high insertion loss and short interaction length the interferometric noise (or etalon-effect) becomes the system resolution limitation. By the insertion of a prism coupled planar waveguide instead of the free-space cell, we demonstrate the concept of integrated optics evanescent wave sensing based on modulation spectroscopy. The system characterization is completed by the validation of the evanescent wave sensitivity in gas of about 8% for a laterally confined waveguide. The outlook of a transcutaneous blood gas monitoring instrument is also established. We first extrapolate the expected waveguide characteristics to show that a resolution of 1mmHg CO2 partial pressure can be reached in 4.5 to 12sec response-time. Then, the requirement of a maximal sensing area of 10mm diameter is at the origin of two geometries of sensor-head: a crossing spiral and a non-crossing double spiral. We verify experimentally a propagation loss as low as 0.5dB/cm and observe a stringent tolerance for light injection. The latter requires an integrated device for mode matching and pigtailing. After implementation of a taper at both waveguide ends, we demonstrate the drastic reduction of the total coupling loss. The carbon dioxide measurements reveal a relatively deep pattern on the second harmonic signal that tends to prove that polarization and resonances are coupled together most likely at the waveguide level. Applying nitrogen (N2) as reference gas, the pattern compensation results in the carbon dioxide signature measurement in direct detection only. The next generation, now in process, is a patented pigtailed non-crossing double-spiral sensor-head building the first step towards an industrial prototype. The waveguide miniaturization allows an interaction length longer than 21cm on a 3×3mm area. The improvement towards a high resolution system concerns the propagation loss and the coupling loss on the sensor-head, and concerns the polarization control or scrambling on the modulation spectroscopy setup. We finally conclude, following the demonstration on the planar waveguide and the extrapolation of the measured characteristics, that the concept of evanescent wave gas sensing based on modulation spectroscopy, can compensate the lack of sensitivity of an integrated optics sensor-head and can be applied in gas and liquid sensing. Moreover, based on the sensor-head developments and realistic improvements, we also conclude that the perspective of high resolution sensing system, and in particular for carbon dioxide determination, evidences the potential applicability of the concept to transcutaneous blood gas monitoring. This opens a new optical outlook of continuous and non-invasive blood gas monitoring, applied in medicine for patient respiratory status monitoring.

EPFL2003

Traffic Anomaly Detection and Diagnosis on the Network Flow Level

Marc Philippe Stöcklin

Monitoring traffic events in computer network has become a critical task for operators to maintain an accurate view of a network's condition, to detect emerging security threats, and to safeguard the availability of resources. Conditions detrimental to a network's performance need to be detected timely and accurately. Such conditions are observed as anomalies in the network traffic and may be caused by malicious attacks, abuse of resources, or failures of mission-critical servers and devices. Behavior-based anomaly detection techniques examine the traffic for patterns that significantly deviate from the network normal activities. Such techniques provide a complementary layer of defense to identify undesired conditions which traditional, signature-based methods fail to detect. These conditions may, for example, emerge from zero-day exploits, outbreaks of new worms, unanticipated user behavior, or deficiencies in the network infrastructure. This thesis is concerned with the challenge of detecting traffic anomalies with behavior-based methods from flow-level network traffic measurements while providing interpretable alert information. We address the problem from two opposite perspectives by analyzing network behavior and individual host behavior. Learning the normal behavior of network activities and detecting relevant deviations thereof is a complex task since behavior changes may also occur under legitimate conditions and should not be reported as anomalies. Due to the absence of explicit detection rules, behavior-based methods moreover provide less precise information as to the causes of aberrant events. Network operators, however, critically depend on meaningful detection results to timely react to alerts by defining effective countermeasures or ruling out potential false alarms. The first part of this work introduces a novel detection scheme which mines for anomalies in the network behavior observed from traffic feature distributions. We study how various types of anomalies may be detected while providing sufficient information to administrators for their characterization. Based on the observation that networks have multiple behavior modes, we propose a method to estimate and model the modes during an unsupervised learning phase. Observed network behavior is compared to the baseline models by means of a two-layered distance computation: Fine-grained anomaly indices indicate suspicious behavior of individual components of traffic features whereas collective anomaly scores for each feature enable effective detection of anomalies that affect multiple components. We show that the two detection layers reliably expose different types of anomalies. Compared with existing detection methods, the resulting alerts provide important additional information that enables administrators to draw early conclusions as to the anomaly causes. In the second part, we address the challenge of processing high-cardinality traffic information in behavior-based anomaly detection. We introduce an adaptive, locality preserving pre-processing method of measurement data into histogram representations with a manageable number of variable-sized bins. Our technique iteratively adapts, with limited, tunable memory requirements, to the empirical distribution of observations in a data stream in order to evenly balance the observations over the histogram bins. As an important result, we show that our method approximates input distributions well and improves the level of detail in histograms compared to traditional methods. Applied to behavior-based anomaly detection, higher detection sensitivity is achieved while, thanks to preserving the locality of observations, the meaning of bins and the interpretability of detection results is retained. In the third part, we focus on the problem that many low-volume anomalies, emerging from individual hosts, are likely to evade from detection because they are not reflected as significant deviations in the variability of aggregate behavior patterns of hosts on the network level. To address this problem, we examine the behavior of individual hosts from their network connection-level activities using an unobtrusive, passive monitoring approach. We develop an unsupervised method to track properties in the activities that recur over time and establish detailed behavior profiles. We propose three anomaly detectors that compare observed activities to the profiles in order to recognize suspicious changes in a host's activities, giving evidence of abnormal behavior. We demonstrate their effectiveness in revealing different types of anomalies, which are not detectable in aggregate network statistics, while providing meaningful alert information to administrators. In addition, we show that the profiles of individual hosts are stable over time and representative of their activities, and may even be used to identify hosts solely from their traffic behavior. In summary, the methods and algorithms presented in this thesis enable practical and interpretable detection of traffic anomalies on the network ow level with behavior-based methods.

EPFL2011

Automatic Red-Eye Removal based on Sclera and Skin Tone Detection

Johann Terrier, Patrick Vandewalle, Flavien Volken

It is well-known that taking portrait photographs with a built in camera may create a red-eye effect. This effect is caused by the light entering the subject’s eye through the pupil and reflecting from the retina back to the sensor. These red eyes are probably one of the most important types of artifacts in portrait pictures. Many different techniques exist for removing these artifacts digitally after image capture. In most of the existing software tools, the user has to select the zone in which the red eye is located. The aim of our method is to automatically detect and correct the red eyes. Our algorithm detects the eye itself by finding the appropriate colors and shapes without input from the user. We use the basic knowledge that an eye is haracterized by its shape and the white color of the sclera. Combining this intuitive approach with the detection of “skin” around the eye, we obtain a higher success rate than most of the tools we tested. Moreover, our algorithm works for any type of skin tone. The main goal of this algorithm is to accurately remove red eyes from a picture, while avoiding false positives completely, which is the biggest problem of camera integrated algorithms or distributed software tools. At the same time, we want to keep the false negative rate as low as possible. We implemented this algorithm in a web-based application to allow people to correct their images online.

2006