DEV Community: Tuna Turacan

A New Cybersecurity Concept: Digital Shadow Security System

Tuna Turacan — Wed, 09 Sep 2026 15:27:56 +0000

An attacker can obtain the password, and even with added security, we can inadvertently grant malicious individuals access. Therefore, a system could be created that learns your behavior within the system over time and see if your behavior within the system aligns with your behavior.

For example, it can learn your typing speed, the times you generally use the system, and your mouse movements, and then display a behavioral match score every time you log in.

People don't behave the same way every day, of course, so the system creates a profile for you over a certain period, and if your behavior doesn't largely match, it requests verification with an authenticator.

If your behavior is not consistent with your previous behavior by 0% to 40%, there will be no restrictions, if the level is between 40% and 70%, continue monitoring and collect additional behavioral signals, if it's between 70-90 percent, sensitive access may be restricted and if the value is even higher, additional verification may be requested directly.

This protection system is an additional system and should not be used alone. Since, user behavior may change over time, someone else may use the user's computer, AI can make behavior mimicry easier and keyboard/mouse data may create privacy problems.

If this project is implemented, password, MFA, digital trust, and shadow security systems can be used together to provide stronger security.

Is Simply Sharing Photos Now A Danger?

Tuna Turacan — Mon, 07 Sep 2026 12:59:51 +0000

Today, billions of people actively use social media. Some people post many photos, videos, and stories publicly. In our digital age, there are many people who are influencers, make a living from social media or are content creators. But how safe is it to constantly share photos with everyone?

Even if we don’t share it constantly or at all, we can be at risk just with our profile picture. This is because deepfake videos can be created very well using only your photo. Sharing just one photo can trigger a chain reaction and put you in a difficult situation.

Nowadays, most photos and videos are not accepted as evidence in any crime unless they are from official or security cameras. This is because the videos that are prepared are so easy and professional. They can do whatever they want with just one photo of yours, publish it, and then demand money to have it removed. They make it so incomprehensible that you might say, "Did I really have a video like that?" If you’re thinking, "Everyone’s sharing their photos, nothing’s happening, so why would they use mine?", well, that’s entirely up to them. Because they don’t need to be hackers or anything like that to create deepfake videos.

In 2024, an employee in Hong Kong was tricked by a deepfake video of the company’s CFO into approving a $25 million wire transfer. Even though this event happened a few years ago, we are now facing far more advanced technology, and privacy is the most important thing protecting us.

Between 2023 and 2024 alone, deepfake scams increased by more than 200% worldwide. According to a report by cybersecurity firm Trend Micro, attackers can now carry out these attacks using free and easily accessible platforms that don’t require expertise.

Don’t share very clear photos of yourself with everyone. When an unknown number calls, don’t speak before the other person can intervene, as your voice could also be used. Perhaps in a few years, people will be afraid to even share their photos, and even the simplest photos will be manipulated and spread on the internet simply because it’s easy to do so.

Moreover, now scammers are creating fictional characters and using them. Criminals create new identities by combining real information belonging to different people. This realistic, AI-generated identity is being used in bank loan applications and money laundering operations. This threat is so significant that, for example, the Saudi Arabian Central Bank has temporarily suspended remote bank account openings after identifying more than 4.8 million suspicious identities.

Unfortunately, there is still no way to protect our photos, audio recordings, and videos other than not sharing them or only sharing them with our closest friends and family. Moreover, 84% of people no longer feel that a convincing video is real evidence. And one in ten people report that fake deepfake images of them have been created without their consent. It has been determined that deepfake photos of young girls are being used for blackmail and revenge purposes even in schools.

From now on, approaching every photo and video with skepticism, not trusting anyone directly, and protecting our privacy are the most important steps we can take. We are entering an era in the digital world where we cannot trust our eyes and ears; therefore, we must find new ways to establish trust.

Instagram Content Moderation: User-level observational study

Tuna Turacan — Thu, 03 Sep 2026 16:20:17 +0000

Observation period: August 2026- September 2026
Duration: 2 weeks

Due to the enormous amount of user-generated content shared on Instagram, effective content moderation is a crucial element for online safety and security. This study adopts a user-level observational approach to examine Instagram's content moderation and reporting mechanisms.

Objective: The aim of this study is to observe how Instagram responds to potentially inappropriate types of content and whether moderation results vary depending on the content's format. The study also aims to examine the effectiveness of Instagram's user reporting mechanism and the feedback provided to users following reports.

Research Questions

1)Does Instagram's observable moderation response differ between images, videos, and direct messages?
2)How often do user reports result in visible enforcement actions?
3)What types of enforcement actions are observed following reports?
4)How consistently does Instagram provide
feedback following a report?

Methodology: Over a period of 2 weeks, I reported over 50 pieces of content shared via Instagram Stories, over 20 Reels videos that I considered potentially inappropriate, and over 20 direct messages that I considered potentially inappropriate. I then observed Instagram's enforcement actions and the feedback it provided.

Fraud/fake accounts: Accounts created for fraudulent purposes and structured to resemble real user profiles were appeared more difficult to identify in observational tests based solely on profile and content. This could increase the importance of user reporting, especially if the account has just been created.

Hate speech: Users are given the option to filter specific words or phrases. However, in the tests conducted, some comments expressing the same meaning with different spellings were able to pass through the filtering mechanism.

Sexual/Inappropriate Content: Rapid intervention was observed in some content that could be clearly deemed inappropriate. Conversely, changes such as reducing the size of the content or censoring certain parts were found to make detection difficult in some cases.

User Reporting System: The effectiveness of user reporting appeared more limited in cases where it could not be automatically determined whether the content constituted a violation.

Feedback Consistency: While the user initially received feedback about the action taken after submitting a report, in some tests no additional notification was received regarding subsequent stages.

Findings: A difference was observed between content formats. Image-based content was found to be subject to moderation more frequently than video content. Only a small percentage of reported Reels videos were removed but some videos were subject to moderation after review. Direct messages, however, were different. In many cases, the notification did not result in an account ban. Instead, temporary messaging restrictions lasting a few days were observed, and in some cases, no visible action was taken.

**Limitations: 1)These observations are based on controlled user-level tests and do not provide information about Instagram's internal moderation models, datasets, or application infrastructure. Therefore, the findings should not be interpreted as a comprehensive evaluation of Instagram's content moderation system.

2)The original content was not available for retrospective analysis. Therefore, the study relies on the observations and moderation outcomes recorded during the observation period.
3)To protect user privacy and avoid unnecessarily reproducing potentially harmful material, original reported content and identifiable information were excluded from the published analysis.**

I hope this study contributes to a better understanding of the user-facing aspects of content moderation and provides useful observations for both users and platform stakeholders.