Levaraging AI Agent (LLM) for OSINT
AI (Artificial Intelligence) is on the rise in the current era, AI has a myriad of benefits and information and makes it easier to do learning, in OSINT we can use this AI for information studies or research and making tools to make your research easier, but it should be noted that information from AI needs to be reviewed and re-learned. There are times when AI gets a lack of information, so my advice is to read journals and papers and news for your basic information
Utilization of AI (LLM)
Data Sciene
AI can be used for data scientists, for example for collecting business information, location mapping and other things. With AI we can do data processing, large-scale modeling such as Spark MLlib or TensorFlow, and can analyze unstructured data, as well as data visualization and other interactive
Sentiment Analysis
AI can do data for sentiment analysis, for example customer reviews from e-commerce platforms, comments on social media, see POIs from people and for business decisions and what I know AI is also multilingual so this is a plus point when you do sentiment for example pros and cons and see emotions in statements and others, for example the TextBlob library
Decision Systems
Decision Support Systems (DSS) and expert systems, when I was in college I studied this, but at that time I didn't really understand it because I didn't focus on this subject XD, from the existing data, for example, we take the example of a disease system in plants with subjects and existing data such as physical characteristics of plants, environment, texture or temperature and others, AI can draw conclusions and process the data and then make decisions. Therefore, this can be used for medical, military and government systems
IoT and Business
In the past, I also played with IoT such as Arduino boards, beacons, Sigfox and Lora. The role of IoT can be integrated with AI for agriculture, for example, smart homes and other things. For example, we take the example of a rice farming system. There is data from rainfall, soil quality, water or irrigation and harvest estimates and an automatic system to care for the rice. Then AI can present data and make decisions automatically. Therefore, AI can detect anomalies or security in the system that you will implement
There are many uses of AI, depending on where you want to implement it and for what, but my advice if you want to use AI, make sure you have basic skills so you don't get confused and AI can explain in detail and clearly without ambiguity, there are many uses of AI in the medical, military, government, information and business fields.
Study Case in OSINT
Language (Inc Slang)
As we know, slang or other countries' languages confuse us and often cause misinformation or miscommunication. In my opinion, AI is quite helpful for us to know the meaning of the language, but sometimes AI doesn't understand it, maybe because of limited data. Therefore, you can use my repo for language OSINT, for example, looking for slang and others. Here is an example.
When I was doing SOCMINT and collecting data from social media, I often found slang and abbreviations that I didn't know. I tried to find the context and found the word "OTEN", for example. I found a netizen explaining the meaning of this word, then I tried to validate it in the GPT chat and it turned out that the GPT chat could find out.
The meaning of OTEN here is "Christian" which is often used by Indonesian netizens to make statements about the Israeli and Palestinian war conflict. If you try scrapping, you will definitely find this word hahaha XD. Here I am not mentioning a particular like racist with religion, it's just an example.
It should be noted that you must know the context or event that you are looking for so that there is no misleading or miscommunication. There is a lot of slang that is intended to be satirical or offensive to organizations or countries, therefore context is key.
Geospatial
Geospatial You can use AI to collect information about geospatial but it depends on the quality of your data collected the more the better, but during my research there were difficulties especially in areas that are not covered by Google Street View or detailed information about the building such as China, Gaza, Papua and others, I tried to use the existing dataset and the articles I created to train this AI to analyze information such as location, buildings, time measurements the results obtained were not very accurate but we can improve this system to be much more accurate, make sure your dataset is correct, good system prompts, integration with MCP and AGI and other things to improve your AI
Here I try to do a comparison with Claude and Mistral and GPT 4o to analyze the image.
The results obtained are not very accurate, but in my opinion this is close to a rough calculation, but if you want details, use time measurement by measuring the direction of light, the sun and objects.
Result from gpt 4o
I have tried with other LLMs like Mistral and deepseek and others the results are not satisfactory some are similar to Claude
Weapon Detection
I tried to do data training with weapon detection, the image above is an airsoft unit not a real unit.
AI can be used to detect weapons, measure the effects of weapons, distance and ammunition used and find out the accessories in the unit. It should be noted for further knowledge I can not explain in more detail, but if you want to try much more detiall try it yourself XD, I tried from the dataset I made and the article I made from the model used like Claude can answer in detail but there are still some that are lacking, I need to update this and do more research, here the detail
Result from gpt 4o
Result from grok
The results obtained are different, the photo unit above is an M4 unit and other specifications such as
Here the deail (original)
M4 unit with stock mk18 and many accesries
ris daniel defense 12" New
stock mk18
rvg
holosight
anpeq laser flashlight
pisir mbus pro metal
In this research, it depends on your system prompt and user prompt. If you are restricted, try doing LLM injection and other methods that I cannot mention in detail.
Data Analysis
Data analysis, AI can be used for data analysis collection such as language, unstructured data and others. It depends on your data I take an example for language
Here I make an AI translator from Indonesian to Chinese
In data analysis, AI can do patterns in the data, for example, AI (especially machine learning) recognizes hidden patterns that are difficult for humans to see from large and unstructured data. In finance, AI can recognize suspicious transaction patterns to detect fraud Or predict
Sentiment Analysis
Sentiment analysis is a Natural Language Processing (NLP) technique used to analyze opinions, emotions, or attitudes in text, whether they are:
Positive
Negative
Neutral
Emontional. Sometimes more specific: angry, happy, scared, etc
AI can process thousands to millions of social media posts, comments, and news articles automatically. Example, AI monitors keywords such as "fuel demonstration", "food prices", "elections" to find out people's attitudes towards certain issues. Sentiment can be used for many things, for example brands, or threat detection or issues, hoax news propaganda, the influence of influencers on young people, the influence of Tiktok social media, stupid trends followed by Indonesian people, here are the steps or summary of collecting information.
Data Collection (scrapping) Role of AI: Collecting data automatically from open sources such as Twitter, Reddit, or forum sites via API or web scraping
Text Cleaning Role of AI: Cleaning text data from spam, emoticons, symbols, and performing language normalization and filtering so that the data is ready for analysis
Sentiment Classification Role of AI: Classifying text into sentiment categories such as positive, negative, or neutral using machine learning or deep learning models
Emotion Detection Role of AI: Detecting specific emotions in text, such as anger, panic, happiness, sadness, using deep learning-based NLP models
Trend & Anomaly Detection Role of AI: Analyzing data patterns to identify spikes in unusual topics or sentiments based on specific times, locations, or contexts
Insight & Reporting Role of AI: Compiling automated analytical reports that can be understood by analysts, journalists, or security officers for fast and accurate decision making
AI Testing LLM or AI for Text Analysis
Not everything can be solved by AI, AI sometimes has errors or wrong information such as miscommunication and other things, therefore it is necessary to be observant in reading text and voice analysis to see more details, but as we know there are many languages from accents and the way people type is completely different not to mention emojis and strange text styles, as for difficult languages such as Chinese or languages that are not alphabetical but characters example (イム, 我, أنا), therefore it is necessary to know a little about the language we will analyze.
Analysis
Well, you can see that there is some strange language (photo 2). If you are not Indonesian, here is a summary from AI.
"Yes, this is in informal Indonesian.
The text expresses a complaint about the media, which the writer believes often highlights unimportant matters (“ilmu tai ayam” or “trivial/pointless knowledge”) instead of religious knowledge. There’s also criticism of capitalism and the media for not giving enough room for religious preaching and truth.
Do you want me to explain the meaning of some confusing words (like “ilmu tai ayam,” “sex pnhafal 30 juz,” and “sesaQ”)?"
Let me explain one by one.
See what I highlighted. If you are not Indonesian, it will be difficult to understand or filter when scraping and doing other things, because there are some ambiguous texts, for example
tay: means “tai” or ‘feces’
bnyk: means “many”
sex: means “once” in formal Indonesian, not “sex” or ‘gender’
nmun: means “however”
sesaq: means “stuffy” or "misguided"
Therefore, language is a challenge for investigators in understanding context, processing information, and intelligence texts. I always mention this language in “ALL About OSINT Matter.” Therefore, if you are a large company or organization conducting a major investigation, my advice is to work with local people or people who understand the language so that there is no miscommunication or misrepresentation of information. But while AI can be useful for analysis, it may take time for it to get better
*Soon i will update......
Last updated