Cloud Hosting

Machine learning and AI continue to reach further into IT services and complement applications developed by software engineers. IT teams need to sharpen their machine learning skills if they want to keep up.

Cloud computing services support an array of functionality needed to build and deploy AI and machine learning applications. In many ways, AI systems are managed much like other software that IT pros are familiar with in the cloud. But just because someone can deploy an application, that does not necessarily mean they can successfully deploy a machine learning model.

While the commonalities may partially smooth the transition, there are significant differences. Members of your IT teams need specific machine learning and AI knowledge, in addition to software engineering skills. Beyond the technological expertise, they also need to understand the cloud tools currently available to support their team's initiatives.

Explore the five machine learning skills IT pros need to successfully use AI in the cloud and get to know the products Amazon, Microsoft and Google offer to support them. There is some overlap in the skill sets, but don't expect one individual to do it all. Put your organization in the best position to utilize cloud-based machine learning by developing a team of people with these skills.

IT pros need to understand data engineering if they want to pursue any type of AI strategy in the cloud. Data engineering is comprised of a broad set of skills that requires data wrangling and workflow development, as well as some knowledge of software architecture.

These different areas of IT expertise can be broken down into different tasks IT pros should be able to accomplish. For example, data wrangling typically involves data source identification, data extraction, data quality assessments, data integration and pipeline development to carry out these operations in a production environment.

Data engineers should be comfortable working with relational databases, NoSQL databases and object storage systems. Python is a popular programming language that can be used with batch and stream processing platforms, like Apache Beam, and distributed computing platforms, such as Apache Spark. Even if you are not an expert Python programmer, having some knowledge of the language will enable you to draw from a broad array of open source tools for data engineering and machine learning.

Data engineering is well supported in all the major clouds. AWS has a full range of services to support data engineering, such as AWS Glue, Amazon Managed Streaming for Apache Kafka (MSK) and various Amazon Kinesis services. AWS Glue is a data catalog and extract, transform and load (ETL) service that includes support for scheduled jobs. MSK is a useful building block for data engineering pipelines, while Kinesis services are especially useful for deploying scalable stream processing pipelines.

Google Cloud Platform offers Cloud Dataflow, a managed Apache Beam service that supports batch and steam processing. For ETL processes, Google Cloud Data Fusion provides a Hadoop-based data integration service. Microsoft Azure also provides several managed data tools, such as Azure Cosmos DB, Data Catalog and Data Lake Analytics, among others.

Machine learning is a well-developed discipline, and you can make a career out of studying and developing machine learning algorithms.

IT teams use the data delivered by engineers to build models and create software that can make recommendations, predict values and classify items. It is important to understand the basics of machine learning technologies, even though much of the model building process is automated in the cloud.

As a model builder, you need to understand the data and business objectives. It's your job to formulate the solution to the problem and understand how it will integrate with existing systems.

Some products on the market include Google's Cloud AutoML, which is a suite of services that help build custom models using structured data as well as images, video and natural language without requiring much understanding of machine learning. Azure offers ML.NET Model Builder in Visual Studio, which provides an interface to build, train and deploy models. Amazon SageMaker is another managed service for building and deploying machine learning models in the cloud.

These tools can choose algorithms, determine which features or attributes in your data are most informative and optimize models using a process known as hyperparameter tuning. These kinds of services have expanded the potential use of machine learning and AI strategies. Just as you do not have to be a mechanical engineer to drive a car, you do not need a graduate degree in machine learning to build effective models.

Algorithms make decisions that directly and significantly impact individuals. For example, financial services use AI to make decisions about credit, which could be unintentionally biased against particular groups of people. This not only has the potential to harm individuals by denying credit but it also puts the financial institution at risk of violating regulations, like the Equal Credit Opportunity Act.

These seemingly menial tasks are imperative to AI and machine learning models. Detecting bias in a model can require savvy statistical and machine learning skills but, as with model building, some of the heavy lifting can be done by machines.

FairML is an open source tool for auditing predictive models that helps developers identify biases in their work. Experience with detecting bias in models can also help inform the data engineering and model building process. Google Cloud leads the market with fairness tools that include the What-If Tool, Fairness Indicators and Explainable AI services.

Part of the model building process is to evaluate how well a machine learning model performs. Classifiers, for example, are evaluated in terms of accuracy, precision and recall. Regression models, such as those that predict the price at which a house will sell, are evaluated by measuring their average error rate.

A model that performs well today may not perform as well in the future. The problem is not that the model is somehow broken, but that the model was trained on data that no longer reflects the world in which it is used. Even without sudden, major events, data drift can occur. It is important to evaluate models and continue to monitor them as long as they are in production.

Services such as Amazon SageMaker, Azure Machine Learning Studio and Google Cloud AutoML include an array of model performance evaluation tools.

Domain knowledge is not specifically a machine learning skill, but it is one of the most important parts of a successful machine learning strategy.

Every industry has a body of knowledge that must be studied in some capacity, especially when building algorithmic decision-makers. Machine learning models are constrained to reflect the data used to train them. Humans with domain knowledge are essential to knowing where to apply AI and to assess its effectiveness.

Read more here:
5 machine learning skills you need in the cloud - TechTarget

A trans-institutional team of Vanderbilt engineering, data science and clinical researchers has developed a novel approach for monitoring bone stress in recreational and professional athletes, with the goal of anticipating and preventing injury. Using machine learning and biomechanical modeling techniques, the researchers built multisensory algorithms that combine data from lightweight, low-profile wearable sensors in shoes to estimate forces on the tibia, or shin bonea common place for runners stress fractures.

The research builds off the researchers2019 study,which found that commercially available wearables do not accurately monitor stress fracture risks.Karl Zelik, assistant professor of mechanical engineering, biomedical engineering and physical medicine and rehabilitation, sought to develop a better technique to solve this problem.Todays wearablesmeasure ground reaction forceshow hard the foot impacts or pushes against the groundto assess injury risks like stress fractures to the leg, Zelik said. While it may seem intuitive to runners and clinicians that the force under your foot causes loading on your leg bones, most of your bone loading is actually from muscle contractions. Its this repetitive loading on the bone that causes wear and tear and increases injury risk to bones, including the tibia.

The article, Combining wearable sensor signals, machine learning and biomechanics to estimate tibial bone force and damage during running was publishedonlinein the journalHuman Movement Scienceon Oct. 22.

The algorithms have resulted in bone force data that is up to four times more accurate than available wearables, and the study found that traditional wearable metrics based on how hard the foot hits the ground may be no more accurate for monitoring tibial bone load than counting steps with a pedometer.

Bones naturally heal themselves, but if the rate of microdamage from repeated bone loading outpaces the rate of tissue healing, there is an increased risk of a stress fracture that can put a runner out of commission for two to three months. Small changes in bone load equate to exponential differences in bone microdamage, said Emily Matijevich, a graduate student and the director of theCenter for Rehabilitation Engineering and Assistive TechnologyMotion Analysis Lab. We have found that 10 percent errors in force estimates cause 100 percent errors in damage estimates. Largely over- or under-estimating the bone damage that results from running has severe consequences for athletes trying to understand their injury risk over time. This highlights why it is so important for us to develop more accurate techniques to monitor bone load and design next-generation wearables. The ultimate goal of this tech is to better understand overuse injury risk factors and then prompt runners to take rest days or modify training before an injury occurs.

The machine learning algorithm leverages the Least Absolute Shrinkage and Selection Operator regression, using a small group of sensors to generate highly accurate bone load estimates, with average errors of less than three percent, while simultaneously identifying the most valuable sensor inputs, saidPeter Volgyesi, a research scientist at the Vanderbilt Institute for Software Integrated Systems. I enjoyed being part of the team.This is a highly practical application of machine learning, markedly demonstrating the power of interdisciplinary collaboration with real-life broader impact.

This research represents a major leap forward in health monitoring capabilities. This innovation is one of the first examples of a wearable technology that is both practical to wear in daily life and can accuratelymonitor forces on and microdamage to musculoskeletal tissues.The team has begun applying similar techniques to monitor low back loading and injury risks, designed for people in occupations that require repetitive lifting and bending. These wearables could track the efficacy of post-injury rehab or inform return-to-play or return-to-work decisions.

We are excited about the potential for this kind of wearable technology to improve assessment, treatment and prevention of other injuries like Achilles tendonitis, heel stress fractures or low back strains, saidMatijevich, the papers corresponding author.The group has filed multiple patents on their invention and is in discussions with wearable tech companies to commercialize these innovations.

This research was funded by National Institutes of Health grant R01EB028105 and the Vanderbilt University Discovery Grant program.

Follow this link:
Vanderbilt trans-institutional team shows how next-gen wearable sensor algorithms powered by machine learning could be key to preventing injuries that...

The Adversarial ML Threat Matrix provides guidelines that help detect and prevent attacks on machine learning systems.

This article is part ofDemystifying AI, a series of posts that (try to) disambiguate the jargon and myths surrounding AI.

With machine learning becoming increasingly popular, one thing that has been worrying experts is the security threats the technology will entail. We are still exploring the possibilities: The breakdown of autonomous driving systems? Inconspicuous theft of sensitive data from deep neural networks? Failure of deep learningbased biometric authentication? Subtle bypass of content moderation algorithms?

Meanwhile, machine learning algorithms have already found their way into critical fields such as finance, health care, and transportation, where security failures can have severe repercussion.

Parallel to the increased adoption of machine learning algorithms in different domains, there has been growing interest in adversarial machine learning, the field of research that explores ways learning algorithms can be compromised.

And now, we finally have a framework to detect and respond to adversarial attacks against machine learning systems. Called the Adversarial ML Threat Matrix, the framework is the result of a joint effort between AI researchers at 13 organizations, including Microsoft, IBM, Nvidia, and MITRE.

While still in early stages, the ML Threat Matrix provides a consolidated view of how malicious actors can take advantage of weaknesses in machine learning algorithms to target organizations that use them. And its key message is that the threat of adversarial machine learning is real and organizations should act now to secure their AI systems.

The Adversarial ML Threat Matrix is presented in the style of ATT&CK, a tried-and-tested framework developed by MITRE to deal with cyber-threats in enterprise networks. ATT&CK provides a table that summarizes different adversarial tactics and the types of techniques that threat actors perform in each area.

Since its inception, ATT&CK has become a popular guide for cybersecurity experts and threat analysts to find weaknesses and speculate on possible attacks. The ATT&CK format of the Adversarial ML Threat Matrix makes it easier for security analysts to understand the threats of machine learning systems. It is also an accessible document for machine learning engineers who might not be deeply acquainted with cybersecurity operations.

Many industries are undergoing digital transformation and will likely adopt machine learning technology as part of service/product offerings, including making high-stakes decisions, Pin-Yu Chen, AI researcher at IBM, told TechTalks in written comments. The notion of system has evolved and become more complicated with the adoption of machine learning and deep learning.

For instance, Chen says, an automated financial loan application recommendation can change from a transparent rule-based system to a black-box neural network-oriented system, which could have considerable implications on how the system can be attacked and secured.

The adversarial threat matrix analysis (i.e., the study) bridges the gap by offering a holistic view of security in emerging ML-based systems, as well as illustrating their causes from traditional means and new risks induce by ML, Chen says.

The Adversarial ML Threat Matrix combines known and documented tactics and techniques used in attacking digital infrastructure with methods that are unique to machine learning systems. Like the original ATT&CK table, each column represents one tactic (or area of activity) such as reconnaissance or model evasion, and each cell represents a specific technique.

For instance, to attack a machine learning system, a malicious actor must first gather information about the underlying model (reconnaissance column). This can be done through the gathering of open-source information (arXiv papers, GitHub repositories, press releases, etc.) or through experimentation with the application programming interface that exposes the model.

Each new type of technology comes with its unique security and privacy implications. For instance, the advent of web applications with database backends introduced the concept SQL injection. Browser scripting languages such as JavaScript ushered in cross-site scripting attacks. The internet of things (IoT) introduced new ways to create botnets and conduct distributed denial of service (DDoS) attacks. Smartphones and mobile apps create new attack vectors for malicious actors and spying agencies.

The security landscape has evolved and continues to develop to address each of these threats. We have anti-malware software, web application firewalls, intrusion detection and prevention systems, DDoS protection solutions, and many more tools to fend off these threats.

For instance, security tools can scan binary executables for the digital fingerprints of malicious payloads, and static analysis can find vulnerabilities in software code. Many platforms such as GitHub and Google App Store already have integrated many of these tools and do a good job at finding security holes in the software they house.

But in adversarial attacks, malicious behavior and vulnerabilities are deeply embedded in the thousands and millions of parameters of deep neural networks, which is both hard to find and beyond the capabilities of current security tools.

Traditional software security usually does not involve the machine learning component because itsa new piece in the growing system, Chen says, adding thatadopting machine learning into the security landscape gives new insights and risk assessment.

The Adversarial ML Threat Matrix comes with a set of case studies of attacks that involve traditional security vulnerabilities, adversarial machine learning, and combinations of both. Whats important is that contrary to the popular belief that adversarial attacks are limited to lab environments, the case studies show that production machine learning system can and have been compromised with adversarial attacks.

For instance, in one case study, the security team at Microsoft Azure used open-source data to gather information about a target machine learning model. They then used a valid account in the server to obtain the machine learning model and its training data. They used this information to find adversarial vulnerabilities in the model and develop attacks against the API that exposed its functionality to the public.

Other case studies show how attackers can compromise various aspect of the machine learning pipeline and the software stack to conduct data poisoning attacks, bypass spam detectors, or force AI systems to reveal confidential information.

The matrix and these case studies can guide analysts in finding weak spots in their software and can guide security tool vendors in creating new tools to protect machine learning systems.

Inspecting a single dimension (machine learning vs traditional software security) only provides an incomplete security analysis of the system as a whole, Chen says. Like the old saying goes: security is only asstrong as its weakest link.

Unfortunately, developers and adopters of machine learning algorithms are not taking the necessary measures to make their models robust against adversarial attacks.

The current development pipeline is merely ensuring a model trained on a training set can generalize well to a test set, while neglecting the fact that the model isoften overconfident about the unseen (out-of-distribution) data or maliciously embbed Trojan patteninthe training set, which offers unintended avenues to evasion attacks and backdoor attacks that an adversary can leverage to control or misguide the deployed model, Chen says. In my view, similar to car model development and manufacturing, a comprehensive in-house collision test for different adversarial treats on an AI model should be the new norm to practice to better understand and mitigate potential security risks.

In his work at IBM Research, Chen has helped develop various methods to detect and patch adversarial vulnerabilities in machine learning models. With the advent Adversarial ML Threat Matrix, the efforts of Chen and other AI and security researchers will put developers in a better position to create secure and robust machine learning systems.

My hope is that with this study, the model developers and machine learning researchers can pay more attention to the security (robustness) aspect of the modeland looking beyond a single performance metric such as accuracy, Chen says.

Read the original post:
The security threat of adversarial machine learning is real - TechTalks