Detecting Social Media Manipulation: Our Investment in Astroscreen
We live in an era where social networks are a core part of our society and have the power to influence some of the most important aspects of our lives. At their outset, Facebook, Twitter, and all the others promised to connect friends and family, enable shared experiences between those physically separated, and even to bring democracy to autocracies. Over time social networks have evolved in ways that shape politics, business, world culture, education, careers, innovation, and more.
In recent years however, we have seen a rise of malicious exploitation of social networks. They are used to spread fake news, turning them into trending topics or public opinion manipulation in insidious ways that are hard to track and prevent. As social networks are designed for virality and reward viral content, once an attack gains steam, it is seen and amplified by real users and covered by real news agencies. Mass manipulation of social media in elections is perhaps the best-known example, just think of the 2016 US elections or the still ongoing Brexit fiasco.
Organized disinformation attacks are supported by fake accounts and automated bot farms. These accounts are inauthentic profiles of nonexistent people built to reach, influence and misinform audiences at massive scale. They are programmed to mimic human-like user behavior by generating messages, following accounts, posting on popular topics, and replying to or sharing particular hashtags. More concerning is that research has shown that a significant portion of users fundamentally believe there is a real person behind each profile.
As bots are becoming increasingly sophisticated (incorporating emerging technologies such as conversational AI and deepfakes), it becomes almost impossible to tell whether they are genuine users. On top of that, as soon as a new technique becomes available to identify these fake accounts and bots, botmakers quickly modify and upgrade their methods to avoid detection. It is a game of cat and mouse that social networks are currently struggling with.
While we know about fake news and social media manipulation in the political space, global brands can just as easily be targeted. As disinformation attacks are becoming more widespread, they are also starting to be recognized as a large and dangerous cybersecurity issue for the corporate world. These coordinated disinformation campaigns can result in costly and damaging consequences for brands, especially iconic ones that have up to 40% of their company’s market value tied up in brand equity. Well-known recent examples of attacks on global brands include the likes of Starbucks and Nike.
Although the widespread problem of social media manipulation has gained significant attention and social networks have started addressing the issue, current techniques of detecting collective anomalies (spammy or automated accounts and behavior in their user base as a whole) have their limitations and are not equipped to detect the sophisticated techniques mentioned above.
This is exactly where Astroscreen fits in. Astroscreen takes an entirely different approach. It uses a mix of machine-learning and human intelligence to detect contextual (instead of collective) anomalies. Its proprietary technology, which is the result of six years of PhD research at UCL, includes techniques such as coordinated activity detection, linguistic fingerprinting and contextual anomaly detection amongst others, and its machine learning classifiers are able to detect bots with a 99% accuracy. Astroscreen’s technology allows them to pinpoint signs of social media manipulation and helps companies to proactively monitor and detect coordinated disinformation campaigns to protect their reputation before any severe damage can be done.
At Speedinvest’s pre-seed practice we focus primarily on the quality of the founders of the companies we invest in. As such, a large part of our conviction is driven by Astroscreen’s founding team, Ali and Juan. They are a passionate, ambitious and resourceful duo tackling an important and widespread problem. Ali (CEO) has extensive experience in building machine learning startups and sold his previous machine learning news analytics company in 2015. Juan (CTO) has a cybersecurity background and completed his PhD on fake account detection on social networks at UCL. In 2017, he exposed the largest Twitter botnet to date, consisting of more than 350,000 automated accounts, that existed undetected since 2013.
Putting all of this together, we believe this is a clear opportunity to build a category-defining company and help companies to counter coordinated disinformation attacks. At Speedinvest, we couldn’t be more excited to be working with the Astroscreen team and we look forward to supporting Astroscreen in identifying and staying one step ahead of social media manipulation.
