Advanced Research

Advanced Research Tools

“Google Dorking” or “Google hacking” is a technique used by newsrooms, investigative organisations, security auditors as well as tech savvy criminals to query various search engines for information hidden on public websites and vulnerabilities exposed by public servers. Dorking is a way of using search engines to their full capacity to penetrate web-based services to depths that are not necessarily visible at first.” (exposingtheinvisible.org).

Dorking can cross ethical and legal boundaries – this is important to keep in mind. However, the use of advanced search operators themselves can help you find information much faster and more efficiently.

This video offers an introduction to advanced search operators, nearby results, cached pages, specific site search, custom range search, stock price search, title search, and other advanced search methods.

See a list of advanced search operators below that you can start experimenting with:

Dork	Description	Google	Duck Duck Go	Yahoo	Bing
cache:[url]	Shows the version of the web page from the search engine’s cache.	✓
related:[url]	Finds web pages that are similar to the specified web page.	✓
site:[url]	Finds pages only within a particular domain and all its subdomains.	✓	✓	✓	✓
inurl:[text]	Finds pages that include a specific keyword as part of their indexed URLs.	✓
allinurl:[text]	Finds pages that include a specific keyword as part of their indexed URLs.		✓
filetype:[file extension]	Searches for specific file types.	✓	✓	✓	✓
intext:[text], allintext:[text], inbody:[text]	Searches text of page. For Bing and Yahoo the query is inbody:[text]. For DuckDuckGo the query is intext:[text]. For Google either intext:[text] or allintext:[text] can be used.	✓	✓	✓	✓
location:[iso code], loc:[iso code], region:[region code]	Search for specific region. For Bing use location:[iso code] or loc:[iso code] and for DuckDuckGo use region:[iso code]. An iso location code is a short code for a country for example, Egypt is eg and USA is us. https://en.wikipedia.org/wiki/ISO_3166-1		✓		✓
contains:[text]	Identifies sites that contain links to filetypes specified (i.e. contains:pdf)				✓

Identifying AI-Generated Content

AI-Generated Text

While generative AI models constantly evolve, you can pay attention to a few things that help you identify what text might be generated by AI. You can use the word “P E A CH (E) S” to help you remember all of them:

P: Punctuation and spelling

Most people make mistakes or inconsistencies in how they spell words and use punctuation. AI models, on the other hand, tend to generate highly polished text in terms of both spelling and punctuation.

E: Expediency

If a single source is producing a ton of content over a short period of time, it can be a sign that they use AI tools for their content output.

A: Accuracy

Checking for certain facts and information in an article can reveal that AI “hallucinated” something false, but did it with casual confidence. Some AI models (such as GPT 3.5) are also only trained on data that doesn’t go past September 2021.

CH: Choice of words

Overly complex jargon without explanation, or much too verbose sentences can be signs of AI-generated text.

H: Habit

If the text has a “habit” of repeating sentence patterns, words, or phrases, it can be a sign of AI-generated text.

S: Style

AI-generated text can lack personality or feelings, making it feel “robotic” or artificial.

Check the two texts in the following columns: Which one do you think is AI-generated?

Consider our era, the age of Artificial Intelligence. Machines are learning and evolving. They mirror us, absorbing the information we provide and imitating our routines.

But true learning goes beyond imitation. It’s about exploring the unknown, embracing change, and seeking novelty. At the heart of learning lies curiosity and the quest for the unfamiliar.

Every new day offers opportunities to learn. Each moment is a call to explore, grow, and learn. Lifelong learning is a joy-filled journey, the magic potion that infuses life with vibrancy.

In life’s grand spectacle, every sunrise, every bird’s song carries a lesson. Lifelong learning is the fuel for a fulfilling life, even amidst the rise of AI.

By embracing lifelong learning, we open ourselves to many opportunities for growth and self-improvement. It instills a sense of curiosity, enabling us to explore new ideas, perspectives, and cultures. It empowers us to challenge the status quo, to question deeply held beliefs, and to evolve as individuals. Lifelong learning fuels our creativity, enhancing our capacity to innovate and adapt to the ever-changing world.

In the age of artificial intelligence, lifelong learning takes on even greater significance. As machines continue to augment our capabilities and redefine the nature of work, our learning ability becomes a vital asset. Lifelong learning equips us with the skills needed to navigate the complexities of the digital age, harness the power of technology, and shape its development in ways that align with our values.

Click for the answer

The first text (on the left) is human, the second text (on the right) is AI-generated

Text excerpts taken from: https://hackernoon.com/ai-generated-vs-human-written-text-technical-analysis

AI-Generated Images

Technology advances constantly, but at the moment, AI makes a few recurring mistakes that allows us to better identify the content it generates. Here are a few of the telltale signs of AI-generated images. (Adapted from https://journal.everypixel.com/how-to-spot-ai-generated-images)

Disproportion

AI can often generate images where things are too big, too small, or just not symmetrical. Keep an eye out for asymmetrical body parts or outsized buildings!

In image 1, the shoulders are clearly lopsided and unnatural.

Image 1: Portrait of a woman with curly hair, a fluffy tiger ears headband and striped face paint

Composition Fusion

When AI combines separate objects or puts things next to each other, the border or transition between these elements can get unexpectedly sudden, or showing a strange fusion.

In image 2, you can check the bottom right corner to see a bird beak merging with a petal.

Image 2: Portrait of a woman with curly hair in a white jacket and an orange hat. Behind her, the wall is filled with orange flower petals and white doves with orange beaks.

Hair Anomalies

The diversity of hairstyles is still hard to nail down for AI algorithms. Make sure to check strange hair patterns, a “glow” around hair, or disconnected hairs.

In image 3, if you check the girl on the right, you can see that some of her hair is growing from her hat.

Image 3: Two women smiling for a photo in an event venue with cozy lighting, wearing witch costumes. Both women have curly hair and tiny cone-shaped witch hats on their heads.

Textual Clues

At the moment, realistic text is hard to create for AI-generated images. This issue might disappear with new models coming out, but until then, you can look out for unreadable, nonsensical, or jumbled text in an image. If fonts, letter sizes, or letter spacing is inconsistent, that can also be a sign of AI, especially in small objects or in the background.

In Image 4, the text on the police officer’s badge is jumbled and not readable.

Image 4: Image of a male police officer in uniform, standing in the middle of a city street.

Hand Positions

Sometimes, AI generates people with unrealistic hand positions, e.g., too many fingers, crooked fingers, or distorted hand shapes. Make sure to check for hand and finger details in images!

In image 5, if you look closely at the fingers on the left hand, you can see their lengths are strange.

Image 5: A businessman in a blue suit sitting at a desk, typing on a laptop.

Perfect Light

When lighting is cinematic or looks very directed and artistic, it can indicate that the image is AI-generated.

In image 6, the lighting situations have complex shadows and a cinematic look, which would be hard to recreate in a real setting.

Image 6: Four images in a 2x2 quadrant. Each image shows a woman and a young boy smiling and sitting on a couch or at a table. — Image 6: Four images in a 2×2 quadrant. Each image shows a woman and a young boy smiling and sitting on a couch or at a table.

Online AI Detection Tools

There are a handful of tools online that say they can detect AI. These tools are not perfect (e.g., they can often show false positives), so they should be used together with other methods, such as reverse image search, geolocation, or shadow analysis.

It is important to note that many cameras already integrate AI features (such as Portrait Mode, Smart HDR, or Night Mode), which blurs the line between “real” and “AI”.

Furthermore, even if no AI signs are detected in an image, the image could still be staged, manipulated, or deliberately framed in a very specific way.

https://www.aiornot.com/

https://hivemoderation.com/ai-generated-content-detection

https://www.v7labs.com/

https://www.invid-project.eu/tools-and-services/invid-verification-plugin/

https://scanner.deepware.ai/

https://app.illuminarty.ai/

https://deepmedia.ai/