Abstract
Artificial Intelligence (AI) has become one of the most polarising topics and eye-catching terms in our contemporary lexicon; seen as either a paragon of modern technology or as a harbinger of humankind’s technological doom, depending on who you ask. From pocket AIs such as Siri to self educating AIs in Silicon Valley, AI has permeated into virtually all facets of life.
Since 2016, US military researchers have invited AI industry experts to the Pentagon to present their capabilities in a formal showcase [1]. Consequently, AI has become inextricable in the Intelligence Community and Intelligence process as a whole. However, despite its omnipresence, AI’s use in the field of Open-Source Intelligence (OSINT) has generally been under-discussed. The purpose of this article is not to discuss the nuts and bolts of AI technology. Instead, it aims to dispel general misconceptions and single out a neglected aspect of the discourse surrounding AI technology as it pertains to Intelligence: OSINT. OSINT refers to the subtype of Intelligence using a multi-factor methodology to collect, analyse and disseminate publicly available information.
By Matthew Sutherland
The Fallacy of Strong AI
Before moving forward, it is important to highlight the notion of Strong AI. Strong AI could theoretically replace human functions, such as analysis or problem-solving. Despite the sensationalization of this notion in media and pop culture, in the contemporary context Strong AI is effectively a pipedream. Experts affirm that “human-like” AI, able to analyze data and provide doctrinal recommendations, have “very little chance in the near future” [2]. Instead, these experts suggest that intelligence analysis remains an art form best handled by experienced human analysts. As such, this article will focus on existing AI technology, and technology in the immediate future.
The Basics of Artificial Intelligence & Big Data
At the most basic level, contemporary AI technology allows for patterns to be identified within data with a high degree of confidence. AI technology can scan massive data sets to identify patterns, anomalies, make associations, and summarize information. These systems are algorithm based and learn from previous information to pull the useful information from these otherwise insurmountable data sets. Machine Learning (ML), the most commonly discussed branch of AI, refers to a field of AI wherein computer algorithms improve through experience. An example of a simple form of ML would be algorithms that analyse the tempo, genre, or instrumentality of a song a user likes to identify similar songs, such as on Spotify.
The proliferation of modern technologies, like social media, has exponentially increased the amount of data that intelligence analysts have to sift through. As technological advancements allow for more intelligence data to be gathered, it is difficult to distil that intelligence into usable tidbits in a timely manner. These massive data sets are referred to as ‘Big Data’. Big Data, simply put, refers to data sets too large for standard computers to handle, meaning they require specialized computing technology. The importance of data to both the private and public sectors have led some to praise Big Data as the ‘world's most valuable resource’. In turn, the processing abilities of AI mean that Big Data and AI tech are inextricably linked [3].
Some Big Data databases take upwards of 24 hours to update due to the sheer volume of intel available. Experts have noted that intelligence apparatus struggle with the unstructured data associated with open-source intelligence, characterizing it as the “biggest” challenge they face today [4]. These massive data sets effectively make identifying important information akin to “finding a specific needle in a stack of needles” [5]. However, this issue can be minimized by AI technology that can comb through these data sets quickly. As technology and intelligence sources continue to proliferate so do the mechanics to effectively engage with the information available.
AI Supported Open-Source Intelligence
Most often, when we discuss AI in an intelligence context it is with an eye towards the formal Intelligence Community, dismissing OSINT applications as being limited to relatively simple functions such as Google Translate or Social Media crawlers. However, as AI technology continues to advance at an exponential pace, so too do its applications for OSINT practitioners.
AI involvement in the OSINT cycle includes disseminating through otherwise insurmountable amounts of data, autonomous imagery, web crawlers, language processing, event detection, pattern identification, etc. Open-source enterprises utilise AI to comb through news from around the world to monitor trends and global situations. AI powered technology has further been used extensively in sentiment analysis for marketing and political campaigns [6] as well as to fact check fake news and identify deep fakes across social media platforms [7]. For example, experts have highlighted Big Data techniques such as web crawling. When combined with a properly selected AI algorithm, web crawling is extremely useful to explore vast sections of social media or other websites in search of specific information or keywords [8].
Maxar Technologies, a space technology company, provides an ML platform for imagery analytics, representing one of the more advanced contemporary uses of AI in the OSINT realm. The Deepcore Suite allows users to train, adapt and deploy ML programs for tagging and validation of satellite, airborne and handheld imagery [9]. It effectively allows the user to consolidate and analyse a much larger set of imagery, the efficiency of which is highlighted by the fact that the program has already been contracted by the United States government [10]. Maxar’s technology is further used for satellite damage assessments after disasters such as wildfires or hurricanes to help direct relief efforts [11]. Maxar is by no means an anomaly in the AI field as numerous companies continue to make leaps and bounds in the field. Tech giants populate the AI field, Google’s publicly available platform TensorFlow utilises AI to analyze the received information and provide predictive answers, prevalent in both Google Translate and Gmail [12]. The adoption of such a technology to the OSINT realm would further streamline the collection and analysis of textual data.
While the application of AI technology is extremely consequential for the defence industry, intelligence and conflict analysis, its use is not all doom and gloom: The Wildlife Conservation Research Unit of Oxford University has utilised very high-resolution satellite imagery and AI algorithms to detect and count African elephant populations with “accuracy as high as human detection capabilities” [13].
The Future of AI Powered OSINT Investigations
The main benefits of current AI use in OSINT boils down to streamlining how data is sifted, sorted, and distributed to be most useful. However, as AI technology continues to advance at an astronomical rate so do its applications for OSINT. Below is a hypothetical scenario highlighting the use of AI in an OSINT investigation [14]:
The process could look like this:
1. The investigator receives highlighted pieces of data from an automated AI algorithm, let’s say this algorithm detects keywords or phrases tagged in Twitter or chatroom posts.
2. They note the presence of a symbol relating to a particular extremist group.
3. Web Crawlers collect information from thousands of sources such as social media, English or foreign language media, news clippings, etc.
4. The collected information is run through pre-determined AI algorithms designed to identify patterns or create a geographic/temporal profile.
5. An intelligence report is then auto-generated according to the consumer’s preferences.
Though none of these instances may themselves seem revolutionary to the intelligence process, the efficiency with which AIs operate significantly streamlines the process and allows for the analyst to spend more time on the analysis, or the “why?”. Therein lies the main benefit of AI technology in the OSINT cycle. The intersection of all these applications of AI in OSINT and the further development of the technology could ideally lead to an AI capable of rapidly identifying pertinent information across all sources and relaying it to the end user, or in other words, creating “Siri with a security clearance” [15].
Conclusion
Despite technological advances, for many OSINT practitioners, crowdsourcing remains the most powerful tool. As organizations, private companies, and hobbyists alike make their investigations publicly available, the collaborative effect galvanises further OSINT investigations, which will continue to synergise with rapidly advancing technology. AI technology continues to be incredibly useful in OSINT for data collection, filtering out misinformation, pattern identification, and for streamlining the overall process. Indeed, some have suggested that AI technology heralds a new dawn for OSINT practitioners, wherein the properly selected algorithm will provide the end user with the exact piece of information they are looking for [16]. Though strong AI capable of replacing human analysis remains a fantasy, the exponential advancements in AI technology mean that such algorithms able to quickly disseminate and identify key pieces of intelligence are anything but a pipedream.
Sources
[1] Keller, J. (2019) Artificial Intelligence Machine Learning Industry Day. Military & Aerospace Electronics. https://www.militaryaerospace.com/computers/article/14069043/artificial-intelligence-machine-learning-industry-day.
[2] Wilson, J.R. (2020) Researchers Set Their Sights on Artificial Intelligence to Coordinate Battlefield Sensors. Military & Aerospace Electronics https://www.militaryaerospace.com/communications/article/16710965/researchers-set-their-sights-on-artificial-intelligence-to-coordinate-battlefield-sensors.
[3] Van Puyvelde, D, Coulthart, S.& Hossain. M. (2017) Beyond the Buzzword: Big Data and National Security Decision-Making. International Affairs 93, no. 6, 1397–1416. https://doi.org/10.1093/ia/iix184.
[4] Ibid
[5] Recorded Future. (2020) How Artificial Intelligence Is Shaping the Future of Open-Source Intelligence. https://www.recordedfuture.com/open-source-intelligence-future/.
[6] Pastor-Galindo, J. et al. (2020) The Not Yet Exploited Goldmine of OSINT: Opportunities, Open Challenges and Future Trends. IEEE Access https://ieeexplore.ieee.org/document/8954668
[7] Altunbey Ozbay. F & Alatas. B (2020) Fake news detection within online social media using supervised artificial intelligence algorithms. Physica A
[8] Recorded Future. (2020) How Artificial Intelligence Is Shaping the Future of Open-Source Intelligence. https://www.recordedfuture.com/open-source-intelligence-future/.
[9] Maxar Technologies. (2021) Deepcore Suite. https://www.maxar.com/products/geospatial-services
[10] Ibid.
[11] Vinton, M. (2020) Maxar Combines High-Resolution Satellite Imagery and Advanced AI/ML Algorithms to Accelerate Building Damage Assessments After Disasters. Maxar Technologies
[12] Maras, M.H. & Alexandrou, A. (2019) Determining authenticity of video evidence in the age of artificial intelligence and in the wake of Deepfake videos. The International Journal of Evidence & Proof, Vol. 23(3) 255–262
[13] Duporge, I. et al. (2020) Using very-high-resolution satellite imagery and deep learning to detect and count African elephants in heterogeneous landscapes. Remote Sensing in Ecology and Conservation.
[14] Recorded Future. (2020) How Artificial Intelligence Is Shaping the Future of Open-Source Intelligence. https://www.recordedfuture.com/open-source-intelligence-future/.
[15] Ibid
[16] Pastor-Galindo, J. et al. (2020) The Not Yet Exploited Goldmine of OSINT: Opportunities, Open Challenges and Future Trends. IEEE Access https://ieeexplore.ieee.org/document/8954668
