Learn the key components and processes of automated forecasting, as well as business use cases, in this 3-part series on time series forecasting.

Unchecked Vulnerability Leaks Information on Millions The headline was almost too big to believe. On Sept 7, The New York Times announced, “Equifax Says Cyberattack May Have Affected 143 Million in the U.S.” This meant that personal credentials, like Security numbers and other data, for almost half the population of the United States was leaked to hackers. The Verge added, “It has been marked as the worst data breach in US history.” As the picture becomes clearer, the issue at stake was a vulnerability in one of the plugins in the Apache Struts framework. Former Equifax CEO Richard Smith said. “It was this unpatched vulnerability that allowed hackers to access personal identifying information.” This week, in Congressional testimony, The Guardian reported, “It’s like the guards at Fort Knox forgot to lock the doors and failed to notice the thieves were emptying the vaults,” Greg Walden, the chairman of the House energy and commerce committee, told Smith. “How does this happen when so much is at stake?” Walden said. “I don’t think we can pass a law that fixes stupid.” The question arises, if Equifax had an AI-powered analytics solution that tracked anomalies in real time,  would this have surfaced the hack immediately, giving the company plenty of time to respond, and thwart any damage? What happened with Equifax? Equifax is one of the three major consumer credit reporting agencies. The company reported on September 7th that hackers had gained access to company data that potentially compromised sensitive information for 143 million American consumers, including Social Security numbers and driver’s license numbers, posing serious repercussions for identity theft. Dan Goodin reported in Ars Technica “The breach Equifax reported Thursday, however, very possibly is the most severe of all for a simple reason: the breath-taking amount of highly sensitive data it handed over to criminals. By providing full names, Social Security numbers, birth dates, addresses, and, in some cases, driver license numbers, it provided most of the information banks, insurance companies, and other businesses use to confirm consumers are who they claim to be.” While still unclear who was behind the attack, with some conjecturing that this was a state-sponsored attack, the data could now be in the hands of hostile governments, criminal gangs, or both and will stay there indefinitely. This leaves over half the population of the US’s vital identifying information exposed. Even worse, while the leak occurred in the spring, the company only went public in September. “The fallout has been swift, with government agencies looking into the incident, class action lawsuits being filed, and consumers demanding free credit freezes.” Why did so much personal data get leaked from Equifax? Cybercriminals exploited a security flaw on the Equifax website. Brian Krebs reported on KrebsonSecurity how the criminals did it. "It took almost no time for them to discover that an online portal designed to let Equifax employees in Argentina manage credit report disputes from consumers in that country was wide open, protected by perhaps the most easy-to-guess password combination ever: “admin/admin.” Looking deeper into the hack, Blogger and admin for SPUZ said, “I asked the hackers one last request before disconnecting. I asked, "How did you manage to get the passwords to some of the databases?" Surely the panels had really bad security but what about the other sections to them? Surely there was encrypted data stored within these large archives no? Yes. There was. But guess where they decided to keep the private keys? Embedded within the panels themselves.” Equifax has confirmed that a web server vulnerability in Apache Struts that it failed to patch months ago was to blame for the data breach. DZone explains how this framework functions. “The Struts framework is typically used to implement HTTP APIs, either supporting RESTful-type APIs or supporting dynamic server-side HTML UI generation. The flaw occurred in the way Struts maps submitted HTML forms to Struts-based, server-side actions/endpoint controllers. These key/value string pairs are mapped Java objects using the OGNL Jakarta framework, which is a dependent library used by the Struts framework. OGNL is a reflection-based library that allows Java objects to be manipulated using string commands.” How could AI-powered Analytics have made an impact? This situation could have been reacted to much faster had the right real-time business intelligence services been integrated into their systems. Approaches like this, such as Anodot’s AI-powered Analytics solution, correlate a company’s raw data to quickly identify anomalous behavior and discover suspicious events in real time, before they become crises. Once an issue is detected technical teams are alerted, so they can resolve issues before they unravel. Companies need to know what their data can tell them right away in order to fix costly problems. Working at the scale of actively monitoring thousands or even millions of metrics, you need an AI-powered analytics solution, with automated real time anomaly detection. Had Anodot’s AI-powered analytics been in place, it could have tracked the number of API Get Requests for user data, and noticed an anomalous spike in requests, catching the breach instantly, regardless of the existing vulnerabilities.

Learn the key components and processes of automated forecasting, as well as business use cases, in this 3-part series on time series forecasting.

Is General Purpose AI Possible? Recently in Forbes, Dan Woods raised the question, ‘Has Anodot Defined the Principles of General-Purpose AI?’ With the rise in attention to Artificial Intelligence (in all its forms), one of the biggest problems that has come up is: knowing when to apply which type of AI to which problem. Woods explained that he has defined a set of principles that can explain when general purpose AI is possible. Can AI Software Really be Good at Everything? So what is general purpose AI anyways? The author established that general purpose AI must be able to solve problems in a wide variety of use cases, be usable by someone who doesn’t really understand the science behind the AI, and should be easy to connect to a business domain. Yet in addressing the practicality of actually applying general purpose AI, the author asserted, "when someone claims their AI is good at everything, it’s usually not that good at anything." While skeptical, he held that one of the main criteria for a successful implementation of general purpose AI would be that the solution can address many business use cases. Ready to Test the General Purpose AI Hypothesis? Woods explained how Anodot’s approach not only peaked his interest, but also intrigued him for how this could test his criteria for realizing general purpose AI. He wrote, “What’s particularly interesting about Anodot (and what makes it more of a general AI product) is that it is data agnostic. Users feed in their data and it helps to diagnose a problem, but it does this without any context – it doesn’t understand the meaning of your data. So, for instance, it doesn’t need to differentiate between sales or social media data to produce a result. It also can process any kind of data, from structured to non-structured.” What’s Anodot’s Secret Sauce? Most traditional analytics solutions may have good collectors (mainly for infrastructure metrics), but fall short when it comes to accurate detection and alerting, with the inevitable result: many legitimate anomalies aren’t detected and too many false positives are reported. Anodot is an Artificial Intelligence analytics solution that finds subtle issues lurking in the data and proactively identifies business incidents in real time. Using an ensemble approach to identify as many anomalies as possible within a given dataset, Anodot runs many algorithms, constantly refining its detection ability while receiving feedback from the user. Its pattern recognition algorithms are designed to detect anomalies in time series data, making it data agnostic. Anodot detects legitimate anomalies with fewer false positives – allowing organizations to remedy urgent problems faster, and capture opportunities sooner. Data-driven Organizations Need Real-Time Business Incident Detection While there are many AI systems in the market that try to find events and data that is not normal, they often fall short. The author asserted, “often these systems fail because they identify too many anomalies (false positives) or not enough (false negatives).” Anodot’s approach differs from other systems, and is ready to match the scale and speed required  by data-driven organizations. “Once these anomalies are identified, a second level of AI evaluates the anomalies, assigning them a score and to correlate them. By using a weighted scale, Anodot tries to surface only major incidents, filtering out low-impact anomalies.” While comprehensively discovering all anomalies, Anodot may not know what the anomalies are and why it could be important that the failure of one component to function impacts the performance of an app. Anodot brings this information to the attention of the user, highlighting a correlation between the two issues. The person using Anodot can quickly turn this information into valuable insights. “This is the key to general purpose AI,” wrote Woods, “The person doing the searching can quickly discern relationships between the anomalies. The model of the anomalies doesn’t have to be captured in the system.” Read the full article on Forbes: Has Anodot Defined the Principles of General-Purpose AI?

A brief overview of the components of today's AI-driven analytics stack, and how they compare to their traditional counterparts. This white paper also surveys the leading solutions for each component.

The claim for most CEO dashboards is that they provide  a complete view of enterprise performance and reliable, real-time information. Yet, if you’ve ever taken the time to read about building the perfect CEO dashboard, you might remember time-consuming tips and tricks explaining which metrics to include in the data monitored by the dashboard, and how that data should be presented. This first step - the selection of which metrics to include - is the fatal flaw of CEO dashboards, because it’s the first opportunity for those who rely on them to miss  critical information. Fatal Flaw #1 CEO dashboards lack intelligent correlation Looking for new business insights and intelligence, and choosing which metrics to include is more art than science.  No one knows the answers to questions that haven’t been asked yet. An important actionable insight can be present in any metric, and why they should all be monitored. More importantly, insights are often only found through correlation of various metrics. One of the keys to making data actionable in any organization is being able to see the whole picture. CEO dashboards often fail to provide all of the necessary information you need to make informed decisions.  These missing links of data can delay a decision, or lead to misinformed decisions, which can be detrimental to your organization. Even if all necessary information is being gathered, it can’t present a coherent picture.  With CEO dashboards, you’re forced to guess what’s important enough to be given the limited real estate on the dashboard. Correlating and acting on this data takes time and manpower, and for larger organizations with a lot of business activity this can add up significant amounts of time before actionable data is consolidated and, if possible, rendered usable. Fatal Flaw #2 Only shows how actual performance meets business goals, but not why CEO dashboards can only indicate how well your company’s actual performance meets your business goals, but can’t show why. If you are lucky enough to benefit from a string of beneficial business events outside your control (social media buzz causes a spike in orders, a competitor suffers a brand-damaging PR mistake, etc.), you won’t ever know that, and more importantly, might get caught unable to respond when that spike hits. Without granular real-time metrics, you won’t connect the cause to the jump in orders. Actionable insight of what caused the spike in orders would allow you to organize enough inventory to respond to the demand, and try to further leverage the buzz for more growth. Fatal Flaw #3 Wasting time driving down the wrong road A CEO dashboard, however, won’t indicate social media buzz until all the hype has died down, if it shows at all. Harvard Business Review  explained that “…dashboards are poor at providing the nuance and context that effective data-driven decision making demands.”  When the bump is over and the top-level KPIs settle back down to normal, you may be identifying a problem to explain the decrease, instead of searching for to leverage the increase – now a lost opportunity. This results in lots of wasted time looking for the wrong root cause, clouding decision-making and leaving your company vulnerable to a competitor’s agile maneuvering. There’s a real cost to relying on dashboards to untangle the correct causation behind a discovered incident. It can even lead to mistaken conclusions, like how a GPS upgrade increased car accidents when it actually significantly decreased them. Fatal Flaw #4 CEO Dashboards don’t provide intelligent prioritization Collecting thousands of events or alerts every minute from your applications and infrastructure, and presenting that data in a dashboard isn’t analytics. The dashboard may look sexy and have beautiful widgets. Users apply filters on this data, performing their own analysis and work. Deriving intelligence from data shouldn’t require an end user to define what to look for, or where, or what are the most critical KPIs, or what normal or abnormal is. This is not intelligence because a user is telling the dashboard exactly what data to show. Fatal Flaw #5 Relevance - interpreting a dashboard a thousand ways CEO Dashboards fail to properly incorporate all of the relevant data sources necessary to make a truly informed, real-time decision, and critical information may not be displayed quickly or effectively enough to act upon.  A single data signal can be both a strong insight for one person and just more noise for another. There’s a level of subjectivity when it comes to the relevance of data. In order to be relevant, data needs to be delivered with the right context, correlation and association. If data isn’t packaged well for decision makers then it will not be acted on. If insights are trapped in a dashboard tool that managers are too overwhelmed to access or the data is delivered too infrequently to use, then the insights may never be found. In business, leaving data up to interpretation can be risky and costly. CEO dashboards belong in the rear-view mirror Business strategy is only effective if you possess the appropriate intelligence and agility to outmaneuver your competition. Dashboards aren’t going to provide insight fast enough when hundreds of thousands of dollars are lost per hour due to a pricing glitch on an ecommerce site, no matter what color scheme, chart type or font is used. Businesses need real time insights This is why CEOs should abandon these dashboards, and turn to an AI analytics platforms to find all the opportunities in your data.

Learn the key components and processes of automated forecasting, as well as business use cases, in this 3-part series on time series forecasting.

South Park Kids Make Jokes and IoT Devices Seriously React In the debut episode for South Park’s 21st season, the beloved characters not only poked fun at the white nationalist movement in an episode called 'White People Renovating Houses', but also yelled commands at their own cartoon Alexa devices throughout the episode. This actually started activating Alexa and Google Home models in the homes of viewers. In the episode, after proclaiming that smart devices are stupid, South Park character Eric Cartman got a smartphone and an Amazon Echo speaker. Cartman’s interaction with Alexa starts innocently enough, with a request to Alexa to set an alarm and tell a joke. The best moments definitely came as the South Park kids asked Amazon Echo’s Alexa to do and say increasingly disgusting things. “It’s a very dumb joke that never gets old because who hasn’t asked their smart home device something idiotic?” This has been seen before. Earlier this year, in Burger King’s “Whopper Burger” television ad, they  activated Google Home devices and had them recite the Wikipedia entry related to Burger King’s Whopper. “The reaction at the time ranged from laughter to outrage. In the end, Burger King and its ad agency won a top industry award, a Cannes Lions for the stunt.” Stirring Up Trouble with IoT Devices Leads to Serious Reactions on Twitter Not only was it revealed that there are people who own both an Alexa and a Google Home, but more seriously that the show could stir up trouble for both. As viewers got further into the episode, some viewers even had to unplug the listening mechanism on their devices. Viewers took to Twitter to share videos and comment on what the episode was doing to their own devices. This episode of South Park drove my Alexa crazy. I never knew my Alexa had such a potty mouth. pic.twitter.com/lWIHySClRd — Tony French (@TonyLFrench) September 14, 2017 New South Park episode is making Alexa go nuts. pic.twitter.com/Rs8oNL7s2u — Tom Buros (@TomBuros) September 14, 2017   Breaking Boundaries Exposes an IoT Weakness Not only was this intrusion a massive violation of entertainment’s fourth wall,  but there is a much greater issue here with the appearance of a new form of ‘digital voice attack’, further weakening our boundaries. “The fact that Alexa doesn’t identify users means that for “her,” all us humans are the same. Actually, you don’t even need a real person to speak.” The world of Internet of Things (IoT) introduces new types problems. While we have felt a general sense of security and control, this act and the targeted technology forces us to redefine our boundaries - borders that are much harder to define and measure because they are not definite. Detecting When Something Goes Wrong with Anodot AI Analytics Anodot actually already warned about the potential for this happening back in February. This raises the underlying question: How do we detect and recognize when these now ubiquitous AI-controlled systems get something wrong? Most AI systems aren’t quite capable of dealing with these difficult, anomalous behaviors. More importantly, this situation could have been avoided, but should serve as a warning to companies making real-time business decisions based on streaming data—there needs to be systems in place to detect when something out of place happens, and to let the right people know. Anodot would have flagged the dramatic spike in search terms and alerted the responsible Amazon team to the issue to investigate further.  Anodot would have been crunching Amazon’s time series data to determine the normal range for these searches. When the spike in Alexa-powered shopping lists rose beyond normal limits, Anodot would have triggered an action and applied a significance score that could help Amazon determine how fast and comprehensive the response might have been.  

Kickstart your business monitoring - see how machine learning anomaly detection can provide your team with the kind of spot-on, real-time alerts that prevent costly incidents and protect revenue.