Machine Learning’s Role in Safe Livestreaming – Security Boulevard

  • Lauren
  • April 14, 2020
  • Comments Off on Machine Learning’s Role in Safe Livestreaming – Security Boulevard

With no concrete end in sight to stay-at-home orders across the globe, people are flocking to unified communications and streaming platforms in droves. Whether they are craving human connection, working remotely or attending classes online, millions are video chatting and livestreaming to maintain business as usual. Unfortunately, the more people who move online, the more likely you are to see them using channels inappropriately. It can be difficult to make security top of mind while keeping up with the rapid speed of increased access. But in today’s world and especially amid this pandemic, creators must put end user safety first if they want to survive.To put this into perspective, consider some of the mainstream collaboration tools that have come under fire recently for hackers spamming their users during private sessions. These breaches have degraded public trust in the safety of these platforms for everything from remote education to work and leisure. For instance, schools across the globe have been grappling with the question of whether education can be securely administered online. Sure, there were already schools that integrated remote learning into their curriculum, but the massive amount of new students who needed to be lifted online during the pandemic opened a door for internet trolls. From hacking into livestreaming virtual classrooms to taking over screens and sharing lewd content, students and teachers have experienced the full gamut of human indecency online in the past couple of weeks.How can these snafus be prevented? Oftentimes, platforms enlist content moderators to prevent abusive content from spamming end users. But, consider the education example again: Public schools typically don’t have such a person on staff. Hiring a “full-time filter” is typically not within the budget of an already strapped-for-cash education system. On top of this, a content moderator position is a uniquely demanding, high-liability job. What happens if a hacker spams an entire classroom with nude photos? Who will be reprimanded? Parents are going to want answers and it will be the school’s job to provide them.This is where content moderation powered by machine learning can solve the most pressing issues surrounding violations with livestreaming and other communications and handle them in a timely manner. By leveraging machine learning technology, images that include anything from nudity to violence can easily be identified and filtered out in real-time. It could no doubt be useful in cases where schools must depend on real-time communications and livestreaming solutions to deliver a safe online learning environment. However, machine learning utility doesn’t end there. Just about every unified communications and streaming platform needs to have content moderation infrastructure in place to be viable in today’s market.Machine Learning and LivestreamingAny application that touts powerful connectivity between users must be prepared for the risks associated with moving masses of people online—and fast. With the pandemic still raging, now is the time to ensure your ducks are in a row. To prevent your app or platform from becoming blacklisted, integrate a machine learning-powered content moderation solution that works for you around the clock.Here are key points to consider if you are wondering about adopting machine learning:People Aren’t Machines“Machines can’t replace human judgment.” This is a common concern but machine learning in the world of content moderation saves people undue stress. Many social media channels invest heavily in contracted content moderators to ensure a safe viewing environment for their end users. For these contracted content moderators, that meant working nonstop to find and flag harmful content before it hit end users—or worse, went viral. That job is anything but easy. Moderators have to watch around 10 to 15 seconds of every questionable video brought to their attention and determine whether it violates the platform’s rules and regulations. Oftentimes this includes content ranging from beheadings to child pornography to animal abuse. As a result, many have endured psychological damage, including PTSD and mental breakdowns.Another glaring issue with the deployment of human content moderators is the fact that the job is often outsourced and underpaid. Huge tech companies hire moderators who are desperate to make a dollar in developing countries. And, with different countries come different cultural sensitivities. More so, what is appropriate in one country isn’t always appropriate in another. It is much easier to plug this type of nuanced data into a machine learning platform versus trying to train a human being to see through the lens of another culture. But because so many of these moderators live abroad, they are put under not only the immense pressure of watching thousands of horrific videos a day but also the anguish of understanding what offends everyone. If they don’t hit the nail on the head every single time, they will likely lose their job. From a developer perspective (and also simply from a place of human decency), it only makes sense to task a non-sentient machine to do this nature of work.Efficiency and Speed is EssentialSo, why is this job a better fit for a computer? The answer is obvious: Computers can work 24/7 mining for abusive content without suffering repercussions for what they “see.” On top of this, they can do it better. Plus, machine learning moderation reduces money spent on training, company liabilities and potential long-term mental damage to employees. Simply put, it’s a win-win. End users are protected and so are the folks working hard to bring them their favorite products.For machine learning to effectively replace human manpower, it needs to be fast and answer the call for public safety. This means the delay between violations and machine decisions needs to be as short as possible while maintaining accuracy. Most cloud-based AI content moderation services have a latency of eight to 10 seconds. But, super-fast solutions can achieve this process in closer to five seconds, from the moment inappropriate content is “seen” by the computer to when it’s taken down. Expecting a human to operate at this speed is impossible, unethical and simply inefficient.Unfortunately, we all know moderating content across the entire internet is, to put it lightly, a lofty task. People sharing inappropriate content, and even bots online, are getting more sophisticated about how they generate and spread content, attempting to co-opt the web to share doctrines or to simply “troll” unsuspecting browsers. Now more than ever, it’s crucial that developers take all necessary actions to monitor the content flowing through livestreaming and other real-time communications technology. If we don’t keep pace with technology as spammers, hackers and negative voices on the internet do, it’s our end users who will suffer the consequences.
Featured eBook
Identifying Web Attack IndicatorsAttackers are always looking for ways into web and mobile applications. The 2019 Verizon Data Breach Investigation Report listed web applications the number ONE vector attackers use when breaching organizations. In this paper, we examine malicious web request patterns for four of the most common web attack methods and show how to gain the context and … Read More #pt-cv-view-e05fc868f1.pt-cv-post-border { margin: 0; border-top-style: solid; border-left-style: solid }
#pt-cv-view-e05fc868f1.pt-cv-post-border .pt-cv-content-item { border-right-style: solid; border-bottom-style: solid; }
#pt-cv-view-e05fc868f1 .pt-cv-title a, #pt-cv-view-e05fc868f1 .panel-title { font-size: 20px !important; line-height: 1.3 !important; font-weight: 600 !important; }
#pt-cv-view-e05fc868f1 .pt-cv-hover-wrapper::before { background-color: rgba(0,0,0,.3) !important; }
#pt-cv-view-e05fc868f1 .pt-cv-content-item:hover .pt-cv-hover-wrapper::before { background-color: rgba(51,51,51,.6) !important; }
#pt-cv-view-e05fc868f1:not(.pt-cv-nohover) .pt-cv-mask * { color: #fff; }
#pt-cv-view-e05fc868f1 .pt-cv-carousel-caption { background-color: rgba(51,51,51,.6) !important; }
#pt-cv-view-e05fc868f1 .pt-cv-specialp { background-color: #CC3333 !important }
#pt-cv-view-e05fc868f1 .pt-cv-specialp * { color: #fff !important; }
#pt-cv-view-e05fc868f1 .pt-cv-pficon { color: #bbb !important; }
#pt-cv-view-e05fc868f1 .pt-cv-custom-fields * { font-weight: 600 !important; }
#pt-cv-view-e05fc868f1 .add_to_cart_button, #pt-cv-view-e05fc868f1 .add_to_cart_button * { color: #ffffff !important; background-color: #00aeef !important; }
#pt-cv-view-e05fc868f1 .woocommerce-onsale { color: #ffffff !important; background-color: #ff5a5f !important; }
#pt-cv-view-e05fc868f1 .pt-cv-readmore:hover { color: #ffffff !important; background-color: #6190bb !important; }
#pt-cv-view-e05fc868f1 + .pt-cv-pagination-wrapper .pt-cv-more , #pt-cv-view-e05fc868f1 + .pt-cv-pagination-wrapper .pagination .active a { color: #ffffff !important; background-color: #00aeef !important; }
[id^=’pt-cv-filter-bar-e05fc868f1′] .active.pt-cv-filter-option, [id^=’pt-cv-filter-bar-e05fc868f1′] .active .pt-cv-filter-option, [id^=’pt-cv-filter-bar-e05fc868f1′] .selected.pt-cv-filter-option, [id^=’pt-cv-filter-bar-e05fc868f1′] .dropdown-toggle { color: #fff !important; background-color: #00aeef !important; }
[id^=’pt-cv-filter-bar-e05fc868f1′] .pt-cv-filter-title { color: #fff !important; background-color: #00aeef !important; }
#pt-cv-gls-e05fc868f1 li a.pt-active { color: #fff !important; background-color: #ff5a5f !important; }
#pt-cv-view-e05fc868f1 .pt-cv-gls-header { color: #fff !important; background-color: #00aeef !important; }

Source: https://securityboulevard.com/2020/04/machine-learnings-role-in-safe-livestreaming/