Identifying sources of light plays an important role in the development of many photonic technologies, such as lidar, remote sensing, and microscopy. Traditionally, identifying light sources as diverse as sunlight, laser radiation, or molecule fluorescence has required millions of measurements, particularly in low-light environments, which limits the realistic implementation of quantum photonic technologies.

In Applied Physics Reviews, from AIP Publishing, researchers demonstrated a smart quantum technology that enables a dramatic reduction in the number of measurements required to identify light sources.

“We trained an artificial neuron with the statistical fluctuations that characterize coherent and thermal light,” said Omar Magana-Loaiza, an author of the paper.

After researchers trained the artificial neuron with light sources, the neuron could identify underlying features associated with specific types of light.

“A single neuron is enough to dramatically reduce the number of measurements needed to identify a light source from millions to less than hundred,” said Chenglong You, a fellow researcher and co-author on the paper.

With fewer measurements, researchers can identify light sources much more quickly, and in certain applications, such as microscopy, they can limit light damage since they don’t have to illuminate the sample nearly as many times when taking measurements.

“If you were doing an imaging experiment with delicate fluorescent molecular complexes, for example, you could reduce the time the sample is exposed to light and minimize any photodamage,” said Roberto de J. León-Montiel, another co-author.

Cryptography is another application where these findings could prove valuable. Typically to generate a key to encrypt an email or message, researchers need to take millions of measurements. “We could speed up the generation of quantum keys for encryption using a similar neuron,” said Magana-Loaiza.

As laser light plays an important role in remote sensing, this work could also enable development of a new family of smart lidar systems with the capability to identify intercepted or modified information reflected from a remote object. Lidar is a remote sensing method that measures distance to a target by illuminating the target with laser light and measuring the reflected light with a sensor.

“The probability of jamming a smart quantum lidar system will be dramatically reduced with our technology,” he said. In addition, the possibility to discriminate lidar photons from environmental light such as sunlight will have important implications for remote sensing at low-light levels.

The post Identifying Light Sources Using Artificial Intelligence appeared first on Artificial Intelligence.

Helsinki: A new learning model for identifying sleep stages with exact precision as that of a physician has been identified by a recent study. The study was conducted by researchers of the University of Eastern Finland.

Sleep is manually classified into five stages, which are wake, rapid eye movement (REM) sleep and three stages of non-REM sleep. The study published in the IEEE Journal of Biomedical and Health Informatics opened up new avenues for the diagnostics and treatment of sleep disorders, including obstructive sleep apnoea (OSA).

OSA is a nocturnal breathing disorder that causes a major burden on public health care systems and national economies. It is estimated that up to one billion people worldwide suffer from OSA, and the number is expected to grow due to population ageing and increased prevalence of obesity. When untreated, OSA increases the risk of cardiovascular diseases and diabetes, among other severe health consequences.

Manual scoring of sleep stages is time-consuming, subjective and costly. To overcome these challenges researchers used polysomnographic recording data from healthy individuals and individuals with suspected OSA to develop an accurate deep learning model for automatic classification of sleep stages.

In addition, the team wanted to find out how the severity of OSA affects classification accuracy. Among healthy individuals, the model was able to identify sleep stages with 83.7 per cent accuracy when using a single frontal electroencephalography channel (EEG), and with 83.9 pc accuracy when supplemented with electrooculogram (EOG).

In patients with suspected OSA, the model achieved accuracies of 82.9 per cent (single EEG channel) and 83.8 per cent (EEG and EOG channels). The single-channel accuracies ranged from 84.5 per cent for individuals without OSA to 76.5 per cent for severe OSA patients.

The accuracies achieved by the model are equivalent to the correspondence between experienced physicians performing manual sleep scoring. However, the model has the benefit of being systematic and always following the same protocol, and conducting the scoring in a matter of seconds.

According to the researchers, deep learning enables automatic sleep staging for suspected OSA patients with high accuracy. The Sleep Technology and Analytics Group, STAG, at the University of Eastern Finland solves sleep diagnostics challenges by using a variety of different approaches.

The methods developed by the group are based on wearable, non-intrusive sensors, better diagnostic parameters and modern computational solutions that are based on artificial intelligence.

The new methods developed by the group are expected to significantly improve OSA severity assessment, promote individualised treatment planning and a more reliable prediction of OSA-related daytime symptoms and comorbidities.

The post New model to identify sleep stages appeared first on Artificial Intelligence.

Businesses today continue to be bombarded by an increasing number of cyberthreats, as hackers become adept at identifying and exploiting vulnerabilities in security systems. A survey by the World Economic Forum ranked data theft and large-scale cyberattacks 4th and 5th in a list of the biggest risks facing our world. With cybercrime regularly hitting the headlines, regulators are implementing new security guidelines and costly fines for violations. Adding to the pressure are consumers who are increasingly prepared to abandon business with a company if they’ve been hit by a data breach. Businesses can’t afford to turn a blind eye to cybersecurity, which has now become a top priority for enterprises.

Attack vs defence: where things stand

The growth of IoT over the last decade has meant that thousands, if not millions, of devices are now contributing to network traffic, and all are potential entry points for attackers. With Gartner predicting that there will be 20.4 billion connected devices by 2020, the potential for unprecedented exposure is only going to continue. Furthermore, the more devices on a network, the more data security analysts have to wade through, making identifying potential threats harder than ever – especially when reports suggest that UK businesses faced a cyberattack every 50 seconds in the second quarter of 2019. While we’re seeing increased awareness around the threat IoT devices can pose, worryingly, cyberattacks on IoT devices have already increased by 300 per cent in 2019.

Compounding the vulnerabilities IoT devices can bring to networks is the nature of cybercriminals, who are constantly evolving their attacks which are becoming increasingly targeted and sophisticated. Furthermore, they’re also collaborating in marketplace environments, sharing tips and advice on how to launch attacks that will cause the most damage.

Most enterprises still rely on traditional approaches to network security to defend against threats. This approach relies on feeding historical data – i.e anomalous activity that was suspicious or malicious – into a learning algorithm so the system knows what to look out for in the future. This enables the system to flag suspicious activity that corresponds to historical data to security teams, and prevent such attacks slipping through the net.

However, this approach is no longer adequate in today’s evolving threat landscape, because it hinders an organisation’s ability to investigate activity that hasn’t been seen before, causing them to miss new attacks. Furthermore, behaviour that is deemed “normal” or “good” within an organisation is constantly evolving, and businesses have to be able to adapt in real time. This legacy approach to network monitoring also places additional stress and burden on security analysts, who don’t have the capacity to sift through the vast amounts of data collected by businesses and identify threats.  It’s no surprise that 56 per cent of senior executives think their cybersecurity analysts are overwhelmed by the sheer volume of data points they need to analyse to detect and prevent threats.

The result? Businesses that can’t identify new and sophisticated attacks, and attackers who are spending an average of 6 months within a network. Clearly, when it comes to enterprise anomaly detection, a change is needed.

Advanced detection: Deep learning & network monitoring

Deep learning powered network monitoring represents a solution to the problem. Increasingly seen as the next generation technology in network monitoring, deep learning is driven by unsupervised algorithms that continuously analyse an organisation’s regular behaviour in order to identify abnormalities. The algorithm is instructed to survey its own infrastructure and proactively search out and unearth the unknown, rather than the known “bad”. This allows businesses to detect unseen threats and take a proactive approach to cybersecurity.

Another advantage of deep learning algorithms is that they have the capability to sift through millions of pieces of data simultaneously in near real-time. The ability to identify anomalous patterns in vast data sets means deep learning network monitoring can perform a level of analysis that’s impossible for humans alone to replicate.

Empowered by deep learning tools, analysts are able to focus on the most rewarding part of their job: the investigation and detection of complex malicious activities. By accelerating access to the information, teams can collaborate and focus on understanding the root cause and the total extent of campaigns against organisations. As a result, security teams’ efficiency is boosted, stress is reduced, cybersecurity analysts’ work is highly valued and the overall organisation security is strengthened.

Businesses can no longer rely on traditional network monitoring methods that provide an inherently binary view of cybersecurity that focuses on good vs. bad behaviour. The volume of data collected by businesses is growing exponentially, and at the same time, cyberthreats are becoming increasingly sophisticated. Add in the fact that cybersecurity teams are under increasing pressure to do more with less and it’s easy to see why enterprises have historically been on the back foot.

Ultimately, deep learning transforms network security from a passive system that is fed seen behaviour, to an active solution that can detect threats in real-time and uncover things not seen before.

The post Boosting enterprise security with deep learning appeared first on Artificial Intelligence.