Orange Tests Deep-Learning Software to Identify Fraud
French multinational telecom operator Orange SA has began testing machine learning software with San Francisco-based startup Skymind to help prevent fraud that occurs on its mobile network. Skymind supports the open source deep learning framework Deeplearning4j and the JVM-based scientific computing library ND4J. Their neural networks are able to perform dimensionality reduction, classification, regression, collaborative filtering, feature learning and topic modeling, often used for enterprise applications related to prediction, data analytics and machine perception.
Orange is using their software to sieve through huge amounts of telephone records to identify changing patterns for instances of fraud. Moving from proof-of-concept to pilot, the goal is to prevent scammers from offering cut rate calling by routing calls over the Internet and illegally using Orange’s mobile network for the last mile of the call, which can ultimately slow network traffic.
However, since scammers can use sophisticated software to evade detection over the network, Skymind’s software is analyzing how often scammers are changing their telephone numbers that can be picked up and abandoned frequently. Once a number is identified as potentially fraudulent, they are then passed onto a human expert for final case determination. The neural network will also learn overtime how to determine fraudulent cases as they are fed more and more raw network data.
Yahoo Reveals How It Live Streamed First Regular Season NFL Game via API
Yahoo demonstrates how good APIs can work at scale with the example of streaming the first regular season NFL game to a globalwide audience. The Buffalo Bills vs Jacksonville Jaguars game reached over 33.6 million total views across all devices on Yahoo and Tumblr. Performance-wise, the streaming had an average rebuffering ratio of nearly 1%, while delivering over 8.5 PB to end users. The stream reached HD levels, with max bit rates over 6.74 Mbps and 60 fps.
Yahoo’s API platform uses Key Performance Indicators (KPIs) that measures latency, availability, error rate, and throughput. The API systems are present in data centers across several regions and the code is deployed with a CI/CD (Continuous Integration/Continuous Delivery) pipeline. A client then proceeds to do a DNS lookup for the API, then issuing a HTTPS request that ends at the Edge to significantly lower latency times; the Edge then forwards this request to the API origin which calls several dependency services, processes the data received, and builds the final response.
As a result, during the 4 hours of the game, the API platform fielded over 215 million calls with a median latency of 11ms, and a 95th percentile latency of 16ms. The APIs showed six 9s of availability during this time period, despite failure of dependencies during spikes in the game.
South Korea Announces $860M AI Fund After AlphaGO Series Game Shock
Responding to the success of Google DeepMind’s Go program AlphaGo winning against grandmaster Lee Sedol in a 4-1 series game, the South Korean government has recently pledged $860 million (1 trillion won) in artificial intelligence (AI) research over the next five years. it does include the founding of an AI research institute with participation from several Korean conglomerates including Samsung, LG Electronics and Hyundai, as well as the technology firm Naver, based near Seoul. The country also plans to set up a separate council for science and technology that will completely revamp its research and development efforts.
The frantic run towards AI development has motivated the country to become technologically competitive. However, there is concern that the fund is going towards industries that could use AI for business applications as opposed to universities who are actually interested in AI research.
GoDaddy Now Selling Cloud Services for SMBs
Small business domain host GoDaddy announced the launch of cloud computing services called Cloud Servers and Cloud Applications. The business model replicates Amazon-style cloud computing services that enables SMBs to build, test and scale cloud solutions on GoDaddy’s shared, flexible infrastructure.
Already hosting more than 10 million websites and giving people software-building tools like WordPress, now their new Cloud Server allows businesses to create raw virtual machines where they can build and run any software on the platform. Along with a poor reputation for guerilla marketing and ridiculous commercial advertising, GoDaddy hopes to tone down their image after going public last year, and offer cloud computing as means to become more of a technology and product provider.
Available in U.S. data centers only, Cloud Server is built on Openstack and powered by Kernel-based VM virtualizations. Cloud applications are powered by open-source library Bitnami for easier deployment, and supported distributions include Ubuntu 14.04, CentOS 6 and 7, Fedora 23, Debian 8, FreeBSD, CoreOS and Arch Linux. Pricing reflects a pay-as-you-go model, starting at $5 per month for the 20GB option. There’s also an 80GB option, capped at $80/month (or $0.12/hour) for 8GB of memory, 4 core processor, 80GB SSD disk and 8TB of transfers.
Kaspersky Warns Steam Stealer Extreme Can Become Lucrative Malware Business
According to Kaspersky Lab researchers, a new malware named Steam Stealer Extreme has opened a new exploits market for as little as $3-$30 to purchase various package builds such as Steam credential rights to a complete user guide with a source code. The Steam Stealer package only requires a user to select their required programming language and comes equipped with an array of APIs and libraries which can be integrated without effort into Steam. With a simple setup at an extremely low cost and the ability to easily bypass firewall and AV protection, Kaspersky warns that the malware application puts over 140 million registered users at risk.
Steam’s multi-OS digital distribution platform provides an always-on cloud environment for users to purchase games and store personal credentials over the system. Steam has become of the most popular gaming platform in the world with more than 12 million active users, but the company is often the target of large-scale attacks. 34,000 users lost their accounts due to the massive DDoS attack last Christmas. This is an ongoing trend of ransomware stealing credentials from gaming sites which may worsen if malware becomes deeply integrated in the next patch updates.