Click Fraud in Advertising: A Comprehensive Study of Mechanisms, Actors, Defense Methods, and Legal Aspects

Abstract

Click fraud is a deceptive practice in search advertising where payment is based on a specific number of clicks on a link. It is a type of cybercrime where contextual advertising ads are clicked by individuals not interested in their content, often using automated scripts or programs. Current analytical data indicates that approximately 10-15% of total ad clicks are click fraud. In 2023, 22% of global advertising spending was lost due to ad fraud, highlighting the significant financial losses incurred by the industry. It is projected that by 2025, losses from click fraud will reach $100 billion, and by 2028, $172 billion, indicating a growing threat to the digital advertising ecosystem.

The evolution of click fraud techniques has progressed from simple automated scripts to highly organized operations involving sophisticated botnets and click farms that mimic human behavior. Initially, fraud relied on basic scripts that repeatedly clicked on ads. However, with technological advancements, malicious actors began using residential proxy networks, machine learning algorithms, and behavior simulation methods to bypass detection systems, making fake clicks virtually indistinguishable from real ones.

Key actors in the click fraud ecosystem include bot and botnet operators, unscrupulous publishers, affiliate programs, and competitors. Their motivations range from direct financial gain obtained through fictitious traffic to targeted sabotage of competitors by depleting their advertising budgets.

To counter click fraud, comprehensive defense methods are employed, ranging from detailed analysis of advertising campaign metrics (such as CTR, bounce rate, IP addresses, and geographical patterns) to the use of specialized software based on artificial intelligence and machine learning. Behavioral analysis, device fingerprinting, and the use of honeypots play an important role. Furthermore, legal regulation and judicial precedents are increasingly playing a significant role in combating this type of fraud.

Effective protection against click fraud and ensuring transparency in the advertising ecosystem requires a comprehensive approach that combines technological innovation, organizational measures, and legal regulation, as well as active cooperation among all market participants.

1. Introduction

1.1. What is Click Fraud and Its Scale

1.1.1. Definition

Click fraud is a deliberate and deceptive practice in digital advertising aimed at manipulating pay-per-click (PPC) models. It is a type of cybercrime where contextual advertising ads are clicked by individuals not interested in their content. The purpose of such actions is to artificially inflate the number of clicks on advertising links, leading to unjustified expenses for advertisers or illicit income for fraudsters. Click fraud can be carried out using automated scripts and programs (bots) or manually, involving human resources.

1.1.2. Scale of the Problem

The problem of click fraud has become global, causing significant damage to the advertising industry. According to analytical data, approximately 10-15% of total ad clicks are click fraud. In 2023, 22% of global advertising spending was lost due to ad fraud, amounting to $84 billion. It is projected that by 2025, global losses from click fraud will reach $100 billion, and by 2028, $172 billion. These figures highlight that the problem is not only persisting but also worsening each year.

The share of non-human traffic on the internet has also significantly increased. In 2023, only 50.4% of global web traffic was generated by real users, while 32% was attributed to malicious bots. This indicates that almost a third of all online traffic is potentially fraudulent or non-targeted. Small and medium-sized businesses are particularly vulnerable to click fraud, losing up to 30% of their advertising budget due to this activity. This may be due to limited resources for detecting and preventing fraud, making them an easy target for malicious actors.

1.1.3. Consequences for Advertisers and the Ecosystem

Click fraud has multifaceted negative impacts on all participants in the digital advertising ecosystem.

Financial Losses: The direct and most obvious consequence is the depletion of advertising budgets. Every fraudulent click spends money that could have been directed towards attracting real, interested users, leading to a decrease in return on investment (ROI). This means companies pay for interactions that do not bring real value or results.

Data Distortion: Fake clicks distort key campaign metrics such as CTR (Click-Through Rate), conversion rate, and average time on site. Distorted data leads to incorrect marketing decisions, as advertisers may mistakenly believe their campaigns are successful and continue to invest in ineffective strategies. This makes it difficult to accurately measure campaign success and optimize strategies.

Loss of Trust: Click fraud undermines trust between advertisers, publishers, and advertising platforms. When advertisers discover that their budgets are being spent on fraudulent traffic, they may reduce spending or completely abandon the use of certain platforms. This creates a cycle where eroded trust leads to investment hesitancy and cautious optimization, which, in turn, creates new opportunities for fraudsters.

Ineffective Retargeting: Bots pretending to be real visitors can be included in retargeting lists. This leads to advertising campaigns spending significant funds on re-engaging non-existent or uninterested users, further depleting the budget and reducing the effectiveness of marketing efforts.

Brand Reputation Damage: If ads are displayed on fraudulent or low-quality websites, it can harm the advertiser’s reputation. Furthermore, ineffective advertising campaigns and low ROI can undermine the trust of stakeholders and partners.

Disruption of User Experience: In some cases, fraudulent activities, such as forced redirects or the display of malicious ads, can degrade the user experience, leading to the installation of ad blockers and avoidance of certain platforms by real users.

Overall, click fraud not only leads to direct financial losses but also destroys the foundation of transparency and effectiveness upon which the digital advertising industry is built.

2. Click Fraud Mechanisms

2.1. Classification and Technical Principles

Click fraud mechanisms are constantly evolving, becoming more sophisticated and difficult to detect. Initially, they were simple scripts, but now they include complex systems that mimic human behavior.

2.1.1. Evolution of Click Fraud Techniques

The evolution of click fraud began with basic scripts that repeatedly clicked on digital advertisements. These early methods were relatively easy to detect due to their repetitive and predictable nature. However, over time, fraudsters developed more sophisticated operations to bypass security systems.

Modern click fraud operations use advanced techniques that make them much more difficult to detect. These include:

• Residential Proxy Networks: These networks allow fraudsters to route their fraudulent traffic through real IP addresses of residential users. This creates the impression that clicks are coming from legitimate users, making detection significantly harder as the traffic blends with genuine user activity.
• Machine Learning: Fraudsters apply machine learning to analyze and imitate the behavior of legitimate users. This enables them to create more convincing fake clicks that are harder for fraud detection systems to distinguish from real ones.
• Behavior Simulation: This technique involves mimicking human-like browsing patterns, mouse movements, page scrolling, and other user interactions. The goal is to make automated clicks appear more authentic, allowing fraudsters to bypass detection mechanisms that look for robotic or unnatural patterns.

These advanced methods enable fraudsters to exploit performance-based advertising models, such as pay-per-click and pay-per-conversion, creating complex challenges for the entire advertising industry.

2.1.2. Types of Click Fraud and Their Technical Principles

Different types of click fraud employ various technical approaches, each with its own level of complexity and detectability.

Simple Clickers: These are automated programs that repeatedly click on a specified screen area at set intervals. They are easy to track due to their repetitive and predictable behavior, ensuring a high level of detection by security systems.

Proxy Bots: These bots use various IP addresses to simulate requests from multiple users, creating the illusion of unique visitors. Modern systems have become more effective at identifying widespread proxy use, so the detection level is medium.

Click Farms: These are physical or virtual organizations where human workers or bots (or a combination thereof) manually perform clicks on specified links for a reward. Physical farms are often located in countries with low labor costs. They can consist of hundreds of smartphones mounted on racks, each connected to multiple fake accounts and constantly updated to avoid detection. Virtual farms use software emulators, proxies, and automation tools to simulate real user behavior. They can mimic device fingerprints, change IP addresses, and interact with applications or websites in complex ways, often bypassing traditional fraud filters. The detection level for such operations varies from low to medium, as human involvement makes them more effective at bypassing automated defense systems.

Sophisticated Bots with Emulation: This is advanced software capable of mimicking human behavior, including mouse movements, random pauses, page scrolling, and viewing multiple pages. Such bots can simulate mouse movements, accept cookies, and even fill out forms, making their detection extremely difficult. They require significant technical expertise for effective implementation, and their detection level is medium-low. Bots can operate alone or as part of larger groups called botnets. Botnets are collections of compromised devices remotely controlled to generate large volumes of fraudulent clicks, significantly increasing the scale of fraud.

Additional methods of click fraud and related fraud include:

• Pixel Stuffing: A deceptive practice where multiple ads are layered on top of each other in a single ad space. When a user clicks on the visible ad, it registers as clicks on all overlaid ads, generating multiple fraudulent clicks from a single interaction. This method significantly inflates click counts without real interaction. It can also involve placing ads in tiny, barely visible formats (e.g., 1×1 pixel) for which advertisers pay, although users do not see them.
• Ad Stacking: Similar to pixel stuffing, where multiple ad creatives are placed on top of each other, and the user only sees one, but all are charged for.
• Domain Spoofing: A form of fraud where a fraudster impersonates the domain of a well-known company to sell low-quality inventory as high-quality. Fraudsters trick buyers into thinking their ads will be shown on a premium site, when in reality they end up on a low-quality website. This leads to advertisers overpaying for ad space that is not valuable. Sophisticated forms of domain spoofing can involve using custom browsers that allow bots to visit any sites and then report a spoofed URL.
• Impression Laundering: Channeling ads through legitimate websites to mask fraudulent impressions.
• Proxy Traffic: Using proxy servers to route traffic through intermediate servers, which masks the true origin of clicks.
• Click Injection: A technique targeting mobile applications that generates clicks that appear to originate from legitimate sources. Often involves manipulating the app installation process to claim credit for installs, thereby earning fraudulent ad revenue.
• Cookie Stuffing: Inserting cookies from other websites into a user’s browser without their knowledge. This manipulates attribution models, allowing fraudsters to claim conversions and earn commissions deceptively.
• Motivated Traffic: Users are offered a reward for performing a target action (e.g., clicking an ad or installing an app). While not always fraudulent, it can distort metrics and not lead to real conversions.
• Organic Site Transitions (Simulation): Simulating live users who enter search engines (Yandex, Google) in incognito mode, type a query, search for a site in the search results, navigate to it, view several pages, mimicking reading, and then leave the site. This is done to improve behavioral factors and the site’s ranking in search results. To enhance the effect, IP addresses can be changed by turning the router off and on.

2.2. Actors and Their Motivations

Click fraud is the result of actions by various actors, each pursuing their own goals, whether financial gain or sabotage.

2.2.1. Competitors

One of the main motivations for competitors is to deplete rivals’ advertising budgets. By repeatedly and fraudulently clicking on a competitor’s PPC ads, they can quickly exhaust their advertising funds. This limits the competitor’s ad visibility and reach, potentially allowing the fraudulent competitor to dominate the ad space and gain an unfair market advantage. The goal is to undermine the effectiveness of the rival’s marketing efforts and reduce their ability to attract real potential customers. Additionally, it aims to increase the cost per click (CPC), making it harder for the competitor to afford their campaigns.

2.2.2. Unscrupulous Publishers and Ad Network Operators

Unscrupulous publishers and other malicious actors in the ad fraud ecosystem are primarily motivated by financial gain. They exploit vulnerabilities in the digital advertising ecosystem to generate fake impressions, clicks, or conversions, thereby stealing advertising budgets without providing real engagement.

Their motivations and impact on advertisers include:

• Budget Depletion: Fraudsters cause advertisers to pay for ads that real users will never see. This means companies spend their advertising budget on fake traffic that does not contribute to sales or brand growth.
• Increased CPA and CPC: Fake interactions generated by bots, for example, inflate the cost per acquisition (CPA) and cost per click (CPC). This makes legitimate advertising campaigns less effective and more expensive for advertisers.
• Profit from Deception: Fraudsters directly profit from advertising spending by manipulating the system. For example, in click fraud, they use bots or click farms to artificially inflate the number of clicks on PPC ads, creating the appearance of good ad performance, while the advertiser spends money on fake engagement.
• Claiming Credit for Conversions: In schemes such as click injection, fraudsters deceive advertisers into believing they were responsible for an app install or other conversion, thereby receiving payment for actions they did not participate in. Similarly, cookie stuffing involves injecting additional cookies to claim credit for sales or leads they did not generate.
• Inflating Ad Space Prices: Through domain spoofing, fraudsters make low-quality websites appear premium, leading to advertisers overpaying for ad inventory that is not valuable.

Overall, the motivation for ad fraud is purely economic for malicious actors who exploit the system for their financial gain at the direct expense of advertisers, undermining the integrity and effectiveness of digital advertising.

2.2.3. Botnet and Click Farm Operators

Botnet and click farm operators are also primarily motivated by financial gain. Click fraud is a multi-billion dollar industry, and with the emergence of each new digital ecosystem, the opportunities for these criminals increase.

• Botnets: Botnet operators use networks of compromised devices to carry out large-scale click fraud operations. Their motivation is to generate revenue through illicit means by mimicking human interactions to avoid detection. These sophisticated bots can perform rapid, repetitive clicks and unnatural browsing sequences, making it difficult for advertisers to distinguish legitimate from fraudulent traffic. By controlling a vast network of devices, botnet operators can generate a huge volume of invalid clicks, leading to significant financial losses for advertisers.
• Click Farms: Click farms are coordinated systems using human workers or bots to simulate user interaction. Their motivation is purely financial, as they are paid to simulate real user traffic. These operations are organized to deceive advertisers into paying for illegitimate clicks, creating the appearance of significant ad engagement, although the clicks do not come from interested potential customers. Click farms are often located in developing countries where cheap labor is readily available.

In addition to financial gain, there are other motives:

• Sabotage: Some hackers may use click bots to harm advertisers without gaining their own financial benefit. This could be an ideological motivation, such as artificially inflating likes or votes to make certain sentiments appear more popular than they actually are.
• Search Engine Ranking Improvement: Cybercriminals may use click fraud to boost the search engine ranking of a malicious web page to make it appear legitimate. The goal in this scenario is to increase the web page’s CTR, thereby improving its search engine ranking and attracting more real users to the page.

Overall, the entire ad fraud ecosystem is driven by the pursuit of illicit profit, with cybercriminals constantly evolving their methods to bypass security measures and exploit new opportunities in digital advertising.

3. Click Fraud Detection Methods

Detecting click fraud is a complex task requiring a multifaceted approach, as fraudsters constantly refine their methods. Modern strategies include data analysis, machine learning, and specialized tools.

3.1. Analysis of Metrics and Behavioral Patterns

Effective click fraud detection begins with careful monitoring and analysis of key traffic metrics and behavioral patterns. Deviations from the norm often indicate fraudulent activity.

3.1.1. Indicators of Fraudulent Traffic

Advertisers should pay attention to a number of “red flags” that may indicate the presence of fraudulent traffic.

• Unusually High CTR (Click-Through Rate) Without a Corresponding Increase in Conversions: If the number of clicks on an ad suddenly increases, but this does not lead to an increase in sales, leads, or other target actions, this is one of the first signs of click fraud. This may mean that someone (or something) is clicking on the ad, but these clicks are not converting into genuine interest or engagement.
• High Bounce Rate and Short Session Duration: If users quickly leave the site after clicking on an ad, without interacting with it or viewing other pages, this indicates a lack of real interest. Bots often exhibit such patterns, as their goal is to generate a click, not to interact with content.
• Click Spikes in a Short Period, Especially from the Same Ad or Keyword Group: Unnaturally fast and repetitive clicks from a single source or on the same ad can be a sign of bot activity.
• Unusual Geographical Patterns: A sudden influx of traffic from regions not targeted, or from countries where the business does not operate, is a serious red flag. Some click farms operate from specific geographical areas, and these locations may show disproportionately high activity.
• Repeated Visits from the Same IP Addresses: High activity from the same IP address, especially with low engagement or conversions, is a strong indicator of fraudulent activity.
• Clicks at Unusual Times of Day: Fraudulent clicks often occur during off-hours, such as very early morning or late at night, when human oversight may be minimal.
• Low Viewability Rate: The Viewability rate determines the actual visibility of an ad. If it is significantly below industry benchmarks (e.g., less than 40-60%), it may indicate that clicks are coming from fraudulent sources, as the ad is not being seen by real people.
• Discrepancy with Industry Benchmarks: If advertising campaign results deviate too much from exemplary metrics for your industry (e.g., clicks, CTR, page views, conversions), this may be a sign of fraud.
• High Traffic, but Low Revenue: While this may indicate website usability issues, it is also a sign of ad fraud when visitors do not convert into buyers.
• Unusual Traffic Sources: For example, traffic from outdated browsers or unpopular device types. Traffic spikes not related to current promotions or seasonal changes in interest can also be suspicious.

Regular monitoring of these patterns helps establish a baseline of normal behavior and makes it easier to identify deviations.

3.1.2. Behavioral Analysis

Behavioral analysis is a powerful tool for detecting ad fraud, as it allows distinguishing between actions of real users and automated or fraudulent interactions. This method is based on machine learning algorithms that continuously learn from new data, making it effective in detecting the constantly evolving methods of fraudsters.

Principles of behavioral analysis:

• Data Collection: Systems collect a wide range of user behavior data, including mouse movements, scroll speed, typing speed, time spent on page, page visit sequence, and click speed. Human users move the cursor randomly and naturally, scroll pages at varying speeds, and click with slight inaccuracy, unlike bots, which exhibit perfectly geometric movements, uniform scroll speeds, and precise, repetitive click positions.
• User Profiling: Based on the collected data, profiles of “normal” behavior for legitimate users are created.²⁵ This allows identifying deviations that may indicate fraudulent activity.
• Monitoring and Pattern Recognition: Behavioral analysis tools continuously monitor user interactions in real-time, identifying any anomalies or patterns characteristic of fraudulent behavior. For example, they can detect unnaturally fast clicks, excessively short sessions, or repetitive actions from the same IP address/device within a short period.
• Risk Assessment and Alerts: Upon detecting a behavioral anomaly, the system assigns it a risk score. A higher risk score indicates a greater likelihood of fraud. This can lead to automatic blocking of suspicious traffic or generation of alerts for further investigation.
• Continuous Learning: Behavioral analysis systems are dynamic and adaptive, constantly updating their algorithms based on new data. This allows them to adapt to new tactics and behaviors of fraudsters, ensuring effective protection against evolving threats.

Examples of behavioral analysis application:

• Time on Page Analysis: Legitimate users spend time reading and interacting with content, while bots often quickly navigate away from a page. Systems can set time-on-page thresholds; overly fast form submissions or ad clicks are automatically flagged as fake.
• Navigation Sequence: Real users follow logical paths when exploring a site, while bots may exhibit random or unnaturally sequential navigation paths.
• Click Precision: Human users tend to click slightly off their intended target, while bots typically click with perfect precision.

Behavioral analysis, especially when combined with machine learning, allows identifying subtle differences in click behavior that can be used to distinguish fraudulent clicks from legitimate ones.

3.1.3. IP Address and Geographical Analysis

IP address and geographical pattern analysis is a fundamental method for detecting click fraud, as fraudulent traffic often has characteristic signs related to its origin.

IP Address Analysis:

• Repeated Clicks from a Single IP Address: Multiple clicks from the same IP address, especially within a short period, are a strong indicator of fraudulent activity. This may indicate the operation of automated bots or click farms.
• Use of VPN/Proxy Servers: Fraudulent clicks often use VPNs or proxy servers to mask the true origin of traffic and bypass detection systems. IP address analysis can reveal the use of such services, which is a suspicious sign.
• IP Address Belonging to Botnets or Data Centers: Traffic originating from data centers, rather than residential or corporate networks where ads are not displayed on the user’s device, indicates the absence of a real human user. Specialized databases, such as Botscout, allow checking IP addresses for botnet affiliation.

Geographical Analysis:

• Unexpected Influx of Clicks from Non-Targeted Regions or Countries: If analytics show a sudden surge in traffic from regions where the business does not operate, or from demographic groups that are not the target audience, this may indicate fraudulent clicks. Click farms often operate from specific geographical areas, and these locations may show disproportionately high activity.
• Geo-Masking: Fraudsters often use proxies or VPNs to mask their true location, making traffic appear to originate from a desired region. This makes it difficult to detect and block fraudulent activity.

Monitoring tools such as Google Analytics and Yandex.Metrica allow tracking these patterns, and specialized solutions like ClickCease and Fraud Blocker provide detailed reports including browser, time, device, location, provider, and keyword information for each click. This allows identifying anomalies and blocking suspicious IP ranges or regions.

3.1.4. Device Fingerprinting

Device Fingerprinting is a sophisticated technique used to identify unique devices interacting with websites or advertisements by analyzing their hardware and software configurations. Unlike cookies, which are easily deleted or spoofed, digital fingerprints are much harder to bypass or hide, making them a powerful tool in combating fraud.

Principle of Operation:

The process of creating a digital fingerprint is typically based on JavaScript code that collects various browser and device attributes when a user first visits a site. This process involves three main steps:

1. Attribute Collection: Data that the device and browser openly provide to establish a connection, as well as other unique characteristics, are collected. Such attributes include:

• IP address.
• Browser type and version.
• Operating system and its details.
• Screen resolution.
• Installed fonts and plugins.
• Time zone settings.
• Graphics Processing Unit (GPU) information and its rendering behavior (WebGL Fingerprinting).
• How the device processes audio signals (Audio Fingerprinting).
• Device battery information.
• HTTP request headers.
• Flash data.
• On-device sensors (gyroscope and proximity sensor).

2. Fingerprint Creation: The collected attributes form a unique identifier using specialized algorithms. Since each device configuration is unique, it is highly unlikely that two devices will have the same digital fingerprint.
3. Fingerprint Verification: On subsequent visits, the system compares the new fingerprint with the stored identifier. Any discrepancies may indicate suspicious behavior, helping the system flag risky devices.

Advantages in Fraud Detection:

• Persistent Identification: Unlike cookies, device fingerprints are much harder to spoof, providing long-term user tracking across multiple sessions and aiding in fraud detection.
• High Accuracy and Uniqueness: Digital fingerprints can uniquely identify users with a high degree of accuracy, as no two devices or browsers behave exactly alike.
• Fraudster Identification: Helps identify suspicious user behavior, including the use of tools and techniques to mask real data and activity.
• Adaptive Fraud Protection: Some solutions use AI and machine learning to continuously evolve, adapting to new fraud tactics in real-time.
• Effective Bot Protection: Digital fingerprints easily identify whether an action was performed by a human or automated scripts by analyzing behavioral patterns and system attributes. This allows detecting coordinated click fraud campaigns and botnets using similar device configurations.

The combination of device fingerprints with contextual signals (e.g., click time, geography, click speed) creates a robust fraud detection system.

3.2. Use of AI and Machine Learning

Artificial Intelligence (AI) and Machine Learning (ML) have become cornerstones in the fight against click fraud, offering advanced capabilities compared to traditional methods. They allow analyzing vast amounts of data in real-time, identifying complex anomalies, and adapting to the constantly changing tactics of fraudsters.

3.2.1. Machine Learning Algorithms for Click Fraud Detection

Traditional fraud detection methods based on blacklists and fixed rules often prove ineffective as fraudsters quickly adapt by changing IP addresses or bypassing static rules. Machine learning overcomes these limitations due to its ability to continuously learn and adapt.

Key algorithms and approaches used in click fraud detection:

• Classification Models: This is the primary approach where a model is trained on labeled data (clicks marked as “legitimate” or “fraudulent”) to predict the type of new clicks.

• Random Forest: Has shown high accuracy (over 95-98.99%) in click fraud detection, demonstrating reliability and generalization ability in various scenarios. It effectively handles imbalanced data, where the number of fraudulent clicks is significantly less than legitimate ones.
• XGBoost, LightGBM, Gradient Boosting: These gradient boosting algorithms also demonstrate very high accuracy (98.90% and above) and are powerful tools for analyzing complex behavioral patterns.
• Decision Tree: Can be used as a standalone classifier or as a basis for ensemble methods.
• Multi-Layer Perceptron (MLP) / Artificial Neural Networks (ANN): Neural networks are capable of identifying complex patterns in large datasets. MLP has shown an accuracy of about 97.34% and high precision in identifying fraudulent clicks (over 98%).
• Logistic Regression and Gaussian Naive Bayes: These algorithms are also applied and can achieve high accuracy (up to 99.78% for logistic regression).
• Anomaly Detection: This approach focuses on identifying unusual patterns or behaviors in datasets that significantly deviate from “normal” behavior.

• Statistical Methods: Use statistical models (e.g., Z-scores, probability distributions) to identify deviations.
• Clustering Methods: Group similar data points and identify outliers that do not fit into any group.
• Proximity-Based Methods: Measure the distance between data points to detect anomalies (e.g., k-Nearest Neighbors (KNN)).
• Isolation Forest: An algorithm that isolates anomalies by randomly selecting features and partitioning data into smaller parts.
• OneClassSVM: An algorithm used for anomaly detection when only data from one class (normal data) is available.
• Recurrent Neural Networks (RNN): Proposed for analyzing temporal click data and effectively distinguishing between genuine and fraudulent actions by learning underlying sequential patterns. RNNs can “remember” previous inputs, which is critical for detecting patterns in click behavior that may span multiple time steps.

Example pseudocode for anomaly detection based on Isolation Forest:

Function DetectClickFraudAnomaly(click_data):
    // click_data: a dataset of clicks containing features such as:
    //   – time_since_last_click
    //   – clicks_per_minute
    //   – ip_address_frequency
    //   – session_duration
    //   – bounce_rate
    //   – geo_location_entropy
    //   – device_fingerprint_uniqueness

    // 1. Data Preprocessing:
    //    – Data cleaning (removing missing values)
    //    – Normalization or scaling of features, if necessary
    //    – Possibly, extraction of additional features (feature engineering)

    // 2. Initialize Isolation Forest model:
    //    – n_estimators: number of trees in the forest (e.g., 100)
    //    – contamination: expected proportion of anomalies in the data (e.g., 0.05 for 5%)
    //    – random_state: for reproducibility of results

    model = IsolationForest(n_estimators=100, contamination=0.05, random_state=42)

    // 3. Train the model on click data:
    //    The model learns to identify “normal” patterns in the data.
    model.fit(click_data)

    // 4. Get anomaly scores for each click:
    //    decision_function returns the anomaly score; the lower, the more anomalous.
    //    predict returns -1 for anomalies, 1 for normal points.
    click_data[‘anomaly_score’] = model.decision_function(click_data)
    click_data[‘is_anomaly’] = model.predict(click_data)

    // 5. Identify anomalous clicks:
    anomalous_clicks = click_data[click_data[‘is_anomaly’] == -1]

    // 6. Return anomalous clicks and their scores
    Return anomalous_clicks

This pseudocode demonstrates how the Isolation Forest algorithm can be used to identify anomalous clicks in a dataset, based on various click and session characteristics.

3.2.2. Role of AI/ML in Real-Time Detection

AI and machine learning play a crucial role in real-time click fraud detection, providing immediate response to fraudulent activities.

• Processing Huge Volumes of Data: Machine learning models are capable of processing vast amounts of advertising traffic data in real-time, identifying anomalies that indicate click fraud. For example, the DataDome system processes over five trillion data signals daily.
• Contextual Detection: Unlike rigid rules, machine learning understands the context of each interaction. It evaluates factors such as device behavior, click frequency, and user engagement patterns to distinguish real users from bots.
• Automatic Blocking: Once an invalid or low-quality user is identified, AI/ML-based systems can automatically prevent ads from being shown to these users, optimizing ad spend and improving overall performance. This allows stopping budget waste before fraudsters get paid.
• Adaptation to New Tactics: Fraudsters constantly change tactics, but machine learning continuously learns from new patterns, blocking threats before they escalate. This ensures that protection remains effective against increasingly sophisticated fraud attempts.
• High Accuracy and Low False Positives: Advanced tools like DataDome claim a false positive rate of less than 0.01%, meaning genuine clicks are rarely mistakenly flagged as fraudulent. Anura also highlights an impressive 99.999% accuracy rate in fraud detection.

AI systems analyze multiple factors, such as click speed and timing, geographical distribution of traffic, unique device IDs and configurations, as well as behavior during user sessions and historical traffic trends. This allows them to identify complex fraud attempts that are difficult to detect manually.

3.3. Technical Protection Mechanisms

In addition to data analysis and the use of AI/ML, there are a number of technical mechanisms that help prevent and detect click fraud.

3.3.1. Honeypots

“Honeypots” are hidden elements on a web page designed to detect and trap bots. They are invisible fields in the form’s code (e.g., an HTML input element hidden with CSS). Since legitimate users do not see these fields, they never interact with them. Bots, however, scanning the HTML code of the page, detect these fields and try to fill them or interact with them. If a hidden field is filled, it is a clear sign that a bot, not a human, interacted with it.

Another example of a “honeypot” is placing two clickable elements, one on top of the other, in the same position on the page. A legitimate user will only be able to click the top element, whereas a bot scanning the page will click both elements. Any attempt to interact with such a hidden element or duplicate clicks is considered suspicious and easily flagged.

The use of “honeypots” helps increase costs for malicious actors. When bots reach “honeypots,” they have to expend computational effort to pass the hidden tests, which increases their CPU time and memory costs, making the attack less profitable and reducing the likelihood of repeated attacks.

3.3.2. JavaScript Challenges and Behavioral Fingerprinting

JavaScript challenges and behavioral fingerprinting are key elements in modern fraud detection systems, especially against sophisticated bots mimicking human behavior.

JavaScript Challenges:

These challenges are lightweight JavaScript code that runs invisibly in the background when a page is visited.11 Bots that cannot render or execute JavaScript are immediately identified. If a visitor fails this test, they are not considered legitimate and are blocked before they can interact with forms or ads. This helps confirm that traffic originates from real browser environments.

Behavioral Fingerprinting:

This method goes beyond simple JavaScript challenges by analyzing subtle yet unique patterns of human behavior that are impossible for bots to replicate at scale. Such patterns include:

• Mouse Movements: Human mouse movements are typically erratic and natural, unlike the perfectly geometric and repetitive movements of bots.
• Scroll Speed: Real users scroll pages at varying speeds, whereas bots may exhibit unnaturally uniform speeds.
• Typing Cadence: The speed and rhythm of typing are also unique to humans.
• Click Speed: The time between clicks, as well as the speed of performing consecutive clicks, can reveal robotic behavior.
• Session Duration: Legitimate users spend varying amounts of time on pages, while bots often exhibit unnaturally short or identical sessions.

Systems using behavioral fingerprinting analyze these signals in real-time. When a visitor behaves like a bot, the system immediately recognizes it and blocks them. This allows distinguishing genuine users from automated attacks, preserving data integrity and protecting budgets.

3.4. Role of Advertising Platforms and Specialized Software

Advertising platforms, such as Google Ads and Meta Ads, have built-in fraud protection mechanisms, but comprehensive protection often requires the use of specialized third-party software.

3.4.1. Built-in Protection Mechanisms of Advertising Platforms

Major advertising platforms invest significant resources in developing and implementing systems for detecting and preventing fraud.

• Google Ads: Google uses machine learning and automatic filters to detect and filter invalid and fraudulent activity. Their automated systems stop advertisers from paying for invalid clicks, for example, if a user immediately leaves the site or if a bot click is detected. Google also manually reviews cases of click fraud that may have been missed by automated systems. Despite these efforts, the Google Display Network (GDN) is often cited as a source of irrelevant or bot traffic. Google Ads has a relatively lower click fraud rate (11% in 2024) compared to some social networks, but higher than Microsoft Ads (14%).
• Meta Ads (Facebook/Instagram): Meta also faces the problem of fraudulent traffic, including “data center traffic,” which is an indicator of bots, and “cyborg accounts” (human-bot hybrids). Despite efforts, it is estimated that at least 5% of Facebook accounts are fake (148 million accounts), many of which are created for fraudulent activity. Meta uses machine learning algorithms and behavioral analysis to identify patterns characteristic of fraudulent clicks.

These built-in systems provide a basic level of protection, but their effectiveness is not absolute.

3.4.2. Specialized Software for Click Fraud Protection

To enhance protection, advertisers often turn to specialized third-party software. These solutions use advanced technologies for more accurate and proactive detection and blocking of fraudulent traffic.

Examples and Capabilities:

• ClickCease: A specialized solution for click fraud protection that blocks bots and suspicious activity. It provides comprehensive user behavior analytics and built-in traffic quality filters. ClickCease automatically blocks suspicious IP addresses in real-time, provides detailed reports on each click (browser, time, device, location, provider, keyword), and allows viewing session recordings to analyze mouse movements.
• Clickfraud: Offers robust technical capabilities for click fraud protection, emphasizing real-time detection, automation, and cross-platform protection.

• Real-time Detection: Uses advanced algorithms to continuously monitor ad campaigns 24/7, instantly identifying and blocking suspicious activity before it depletes the budget.
• Automation: Actively blocks bots and malicious competitors, eliminating the need for manual intervention.
• Cross-platform Protection: Provides comprehensive protection across major advertising platforms, including Google Ads, Microsoft Ads (Bing), and Meta Ads (Facebook/Instagram). For Google Ads, it protects various campaign types, including Performance Max (PMax), ensuring budget is directed to real potential customers. For Meta Ads, it includes “Bot Probability Assessment” and customizable protection controls.
• Reporting: Provides automated reports with deep insights into campaign performance, keyword effectiveness, and click fraud threats.

• Anura: Positioned as a highly accurate and reliable way to stop click fraud. Uses behavioral analysis, device fingerprinting, and IP reputation monitoring. Anura automatically hides ads from fraudulent visitors, preventing budget waste. Claims 99.999% accuracy in identifying unwanted visitors.
• TrafficGuard: Uses tools to block fake clicks on Meta campaigns, improving ROI. Proactively prevents ads from being shown to invalid traffic sources through continuous monitoring and click fraud prevention. Provides “intelligent real-time analytics” and AI/ML protection for social networks.
• HUMAN Security (Ad Click Defense): A solution designed for advertising platforms to protect revenue and maintain advertiser trust by detecting and filtering invalid clicks. Uses an extensive network (20 trillion digital interactions per week, 3 billion unique devices per month) and over 400 detection algorithms. Provides real-time detection of sophisticated (SIVT) and general (GIVT) invalid click traffic.

General Technical Features of Specialized Software:

• AI and Machine Learning: A central element that allows analyzing vast amounts of data, identifying complex patterns, and adapting to new threats.
• IP Blocking and Geo-filtering: Automatic exclusion of known fraudulent IP addresses and restriction of ad display in high-risk geographical regions.
• Device Fingerprinting: Creation of unique digital signatures for each visitor to identify suspicious activity and mask the identity of fraudsters.
• Behavioral Analysis: Tracking mouse movements, scroll speed, time on site, and other behavioral signals to distinguish humans from bots.
• Click Thresholds: Setting limits on the number of clicks from a single IP address or device within a certain period.
• Blacklists and Whitelists: Managing lists of suspicious and trusted traffic sources.
• Reporting and Analytics: Providing detailed reports on traffic quality, detected fraud, and performance metrics.

These tools allow advertisers not only to detect but also to actively prevent fraud, ensuring that their advertising budget is spent on real, interested users.

4. Click Fraud Protection Methods

Effective click fraud protection requires a comprehensive approach that combines internal monitoring, the use of specialized tools, and strategic planning.

4.1. Recommendations for Advertisers

Advertisers can take a number of specific steps to minimize damage from click fraud.

4.1.1. Internal Monitoring and Data Analysis

Regular and detailed analysis of advertising campaign data is the first line of defense.

• Establishing Baselines and Benchmarks: Define what constitutes “normal” behavior for your campaign and track key metrics such as CTR, bounce rate, session duration, and conversion rate. This will help quickly identify deviations.
• Traffic Monitoring: Closely monitor sudden traffic spikes, especially during off-hours or from unexpected geographical regions. A discrepancy between high click volume and low conversion rates is a strong indicator of fraud.
• IP Address and Geographical Pattern Analysis: Look for suspicious patterns, such as high activity from the same IP address with low engagement or conversions. Geographical analysis can reveal an influx of clicks from regions not targeted.
• Monitoring Click Time and Frequency: Clicks occurring at uniform intervals or clusters of clicks within a few seconds are likely bot activity.
• Using Web Analytics Tools: Google Analytics 4 (GA4) and Yandex.Metrica offer built-in features for detecting bot traffic, including web scrapers and known malicious sources. GA4 allows using trend change and anomaly detection, which automatically alert when data falls outside expected norms.

4.1.2. Ad Campaign Settings and Targeting

Optimizing campaign settings and targeting can significantly reduce vulnerability to click fraud.

• IP Address Exclusion and Geo-targeting: If fraudulent IP addresses are identified, they can be blocked in the ad platform settings. Geo-targeting allows directing ads to specific regions, excluding high-risk areas.
• Ad Scheduling: Running ads only during peak activity hours of the target audience can reduce the impact of click bots, which often operate outside normal viewing hours.
• Audience Refinement and Remarketing: Configuring ad display only to an audience genuinely interested in the product reduces the likelihood of click fraud. Remarketing focuses ad spend on more qualified, previously engaged users.
• Using Negative Keywords: Adding irrelevant keywords as negative keywords prevents ads from being shown for queries that may attract bot traffic or uninterested users.
• Conversion Tracking Setup: A crucial step for understanding the real effectiveness of advertising. If many clicks are observed but few conversions, this is a clear sign of click fraud. Integrating a CRM system with advertising platforms allows training algorithms to find higher-quality leads, excluding spam ones.

4.1.3. Monitoring and Protection Tools

For more advanced protection, it is recommended to use specialized tools.

Using such tools allows automating the detection and blocking of fraudulent traffic, providing significant budget savings and increased ROI.

4.1.4. Collaboration with Advertising Platforms and Legal Measures

In case of serious click fraud detection, it is important to interact with advertising platforms and consider legal actions.

• Reports and Complaints to Advertising Platforms: Google and Yandex accept complaints about suspicious traffic. Providing evidence (logs, analytics reports) can help in reviewing charged funds.
• Collaboration with Partners: Choose advertising networks and platforms with robust fraud detection and prevention mechanisms. Demand transparency and detailed reports on ad placement and interaction.
• Legal Consultation: In case of significant damage from fraud, an advertiser may resort to legal action. It is important to understand applicable legislation and possibilities for legal protection.

4.2. Legal Aspects of Click Fraud

Click fraud, as a form of fraud in the field of computer information, falls under various laws and can have serious legal consequences.

4.2.1. Legislation on Ad Fraud and Cybercrime

Click fraud can be classified as a crime in various jurisdictions.

• Russian Federation:

• Article 159.6 of the Criminal Code of the Russian Federation “Fraud in the Sphere of Computer Information”: Defines fraud as the theft of another’s property or the acquisition of rights to another’s property by entering, deleting, blocking, modifying computer information or otherwise interfering with the functioning of means of storing, processing, or transmitting computer information or information and telecommunication networks. Click fraud, which involves generating false clicks to deplete an advertiser’s budget or increase a publisher’s earnings, may fall under this article, as it involves “entering” or “modifying computer information” (click data) and “interfering with the functioning of information and telecommunication networks” (advertising platforms). Penalties range from fines to imprisonment for up to ten years, depending on the scale and circumstances of the crime (e.g., committed by a group of persons, using official position, especially large scale).
• Article 14.33 of the Code of Administrative Offenses of the Russian Federation “Unfair Competition”: For unfair competition, including competitor fraud in contextual advertising, fines are provided: for legal entities from 100,000 to 500,000 ₽, for individual entrepreneurs from 12,000 to 20,000 ₽.
• Article 1515 of the Civil Code of the Russian Federation “Liability for Illegal Use of a Trademark”: For using another’s trademark, the infringer pays compensation from 10,000 to 5,000,000 ₽, the exact amount is determined by the court.
• Federal Law No. 38-FZ of March 13, 2006 “On Advertising”: Regulates what can and cannot be done in advertising. Prohibits disparaging competitors, using offensive images, misleading consumers, incorrect comparisons. Violations lead to fines under Article 14.3 of the Code of Administrative Offenses of the Russian Federation.
• United States of America:

• Computer Fraud and Abuse Act (CFAA) (18 U.S.C. § 1030): The primary federal law used to prosecute hackers and other cybercriminals. It covers unauthorized access to computers, data theft, malware distribution, and network intrusions. In Juju, Inc. v. Native Media, LLC (2020), it was established that click fraud violates the CFAA, which became an important legal precedent. To be charged under the CFAA, it is required to prove unauthorized access, intent to defraud, and causing damage.
• Wire Fraud Statute (18 U.S.C. § 1343): Criminalizes any scheme to defraud involving “electronic communications.” Click fraud can be prosecuted under this law if it involves deception and financial loss.
• Federal Trade Commission (FTC): Regulates online advertising practices and can take action against fraudulent activities. The FTC files lawsuits in federal courts to stop fraud, prevent future schemes, freeze assets, and obtain compensation for victims.
• European Union:

• General Data Protection Regulation (GDPR): While primarily focused on data protection, it also covers issues related to the misuse of personal data, which may include data used for fraudulent activities such as click fraud. Violations can lead to significant fines.
• Unfair Commercial Practices Directive (UCPD): Prohibits unfair business practices, including deceptive and fraudulent acts that distort consumer behavior. Click fraud can be prosecuted under this directive if it is proven to deceive or mislead consumers.
• United Kingdom:

• Fraud Act 2006: Makes it illegal to commit fraud by false representation, failure to disclose information, or abuse of position.
• Computer Misuse Act 1990: Criminalizes unauthorized access to computer systems.
• Other Countries: Legislation in other countries (e.g., Australia, Canada, India, Japan) also contains provisions against cybercrime, fraud, and unfair competition that can be applied to click fraud.

4.2.2. Judicial Precedents and Challenges in Proving

Court cases on click fraud demonstrate the complexity of proving and the need for reliable evidence.

• Google v. Auction Experts: Google won a lawsuit against Texas-based company Auction Experts (a publisher), which it accused of paying people to click on ads on its website, costing advertisers $50,000. This case highlights that even large platforms actively combat fraud.
• Uber Lawsuits: In December 2019, Uber filed a lawsuit against five ad networks, alleging they falsified reports to conceal fraudulent ad operations. Uber claimed that two-thirds of its ad budget ($100 million) was wasted on low-quality or fraudulent ads that did not result in conversions. In January 2021, Uber won a separate lawsuit against Phunware Inc., where it was proven that a significant number of mobile app installs were faked using “click flooding.” Phunware also placed Uber ads on pornographic sites, which was a breach of contract.
• Criteo v. SteelHouse Lawsuit: Ad network Criteo filed a lawsuit alleging that rival firm SteelHouse (now MNTM) used a “fake click fraud scheme.” Criteo claimed that SteelHouse used “bots and other automated means” to generate fake clicks, leading to client loss.

Challenges in Proving Click Fraud:

Proving click fraud is one of the most difficult tasks in the digital age.

• Identification: Distinguishing fraudulent clicks from legitimate ones is extremely difficult, as fraudsters mimic human behavior.
• Attribution: Determining the source of fraudulent clicks is difficult, especially when click fraud is carried out using bot farms or “fraud-as-a-service” tools.
• Cross-border Nature: Fraudsters often operate from different countries, making prosecution difficult.
• Lack of Direct Legislation: In some jurisdictions, the activities of click farms are in a “gray area” and not directly regulated, although they may violate consumer protection or unfair competition laws.

The use of blockchain technology can help in providing digital evidence by providing an immutable and transparent record of transactions and actions. In one case, blockchain helped recover $120,000 in click fraud damages.

5. Conclusion and Recommendations

Click fraud represents a serious and constantly evolving threat to the digital advertising industry, leading to billions in losses annually and undermining trust within the ecosystem. The evolution of fraudulent techniques from simple scripts to sophisticated botnets and click farms mimicking human behavior requires advertisers and platforms to constantly adapt and implement advanced protection methods.

The main actors in click fraud—competitors, unscrupulous publishers, and botnet/click farm operators—are driven by financial gain or the desire to harm rivals, which underscores the need for a multifaceted approach to counteraction.

Effective click fraud protection cannot rely on a single method. It requires a comprehensive strategy, including:

1. Continuous Monitoring and In-depth Data Analysis: Advertisers should actively use analytical tools such as Google Analytics and Yandex.Metrica to identify traffic anomalies, such as unusually high CTR without conversions, short sessions, traffic spikes from non-targeted regions, or repetitive clicks from the same IP addresses. Establishing baselines and regular campaign audits are critically important.
2. Implementation of Advanced Technical Solutions: The use of specialized software for click fraud protection (e.g., ClickCease, ClickGUARD, Anura, TrafficGuard, HUMAN Security) is essential. These AI and machine learning-based solutions can analyze behavioral patterns, device fingerprints, IP addresses, and other signals in real-time to identify and block fraudulent traffic. The use of “honeypots” and JavaScript challenges also increases the effectiveness of bot detection.
3. Optimization of Ad Campaign Settings: Careful targeting by location and audience, the use of negative keywords, and strategic scheduling of ad displays can significantly reduce vulnerability. A key aspect is also the setup and continuous monitoring of conversion tracking so that advertising platforms learn from real, not fraudulent, leads.
4. Collaboration with Advertising Platforms: Regular interaction with the technical support of Google Ads, Meta Ads, and other platforms, providing them with evidence of fraud, and requesting a review of charged funds can help in recovering damages and improving built-in protection systems.
5. Legal Literacy and Readiness for Legal Action: Understanding applicable legislation (such as Article 159.6 of the Criminal Code of the Russian Federation, CFAA, unfair competition laws) and judicial precedents allows advertisers to assess risks and opportunities for legal protection. In the case of large-scale fraud, legal action may become a necessary step, with the collection and securing of digital evidence (possibly using blockchain technologies) being crucial.

In conclusion, combating click fraud is an ongoing process that requires investment in technology, staff training, and active cooperation among all participants in the digital advertising ecosystem. Only a comprehensive and proactive approach will protect advertising budgets, ensure data accuracy, and maintain the trust necessary for the healthy development of online advertising.

Источники

1. что это, как распознать и бороться с кликфродом – Drozd.red, дата последнего обращения: июня 4, 2025, https://drozd.red/glossary/klikfrod/
2. Кликфрод (click fraud) – мошеннический клик – SEOnews, дата последнего обращения: июня 4, 2025, https://m.seonews.ru/glossary/klikfrod-/
3. Ad Fraud Impact 2025 – Tapper, дата последнего обращения: июня 4, 2025, https://www.tapper.ai/ad-fraud-impact-2025
4. PPC Statistics: 40 Key Trends and Insights To Optimize Paid Advertising Campaigns in 2025, дата последнего обращения: июня 4, 2025, https://www.designrush.com/agency/paid-media-pay-per-click/trends/ppc-statistics
5. Click Farms Create Fake Likes and Subscribers. SEE HOW! – YouTube, дата последнего обращения: июня 4, 2025, https://www.youtube.com/watch?v=kZoB7FnkGk8
6. 2024 Click Fraud Statistics: Ad Spend Impact & Prevention – Twinleon, дата последнего обращения: июня 4, 2025, https://www.twinleon.com/blog/en/click-fraud-statistics-2024/
7. Click fraud: Understanding the impact on digital advertising – HUMAN Security, дата последнего обращения: июня 4, 2025, https://www.humansecurity.com/learn/blog/click-fraud-understanding-the-impact-on-digital-advertising/
8. What is Click Fraud & How to Prevent Click Bots | Lotame, дата последнего обращения: июня 4, 2025, https://www.lotame.com/resources/putting-end-targeting-bots-click-fraud-prevent/
9. What Is a Click Farm? Spot and Block Fake Clicks Fast| Spider Af …, дата последнего обращения: июня 4, 2025, https://spideraf.com/articles/what-is-a-click-farm-how-to-spot-and-stop-fake-clicks
10. Ad Fraud Guide: What It Is, How It Works, And Main Types | Adtelligent, дата последнего обращения: июня 4, 2025, https://adtelligent.com/blog/ad-fraud-guide-what-it-is-how-it-works/
11. Click Fraud – Risks & Prevention Strategies | Indusface, дата последнего обращения: июня 4, 2025, https://www.indusface.com/learning/click-fraud/?amp
12. Why click fraud is costing your business – and how to stop it – Adido …, дата последнего обращения: июня 4, 2025, https://www.adido-digital.co.uk/blog/why-click-fraud-is-costing-your-business-and-how-to-stop-it/
13. Click Bots 101: 10 Must-Know Tips For Effective Prevention And Detection, дата последнего обращения: июня 4, 2025, https://www.wisegrowthmarketing.com/click-bots/
14. 10 Proven Strategies for Click Fraud Protection in Digital Advertising – Anstrex, дата последнего обращения: июня 4, 2025, https://www.anstrex.com/blog/10-proven-strategies-for-click-fraud-protection-in-digital-advertising
15. What Are Ad Fraud Solutions and Why Are They Important? – Anura, дата последнего обращения: июня 4, 2025, https://www.anura.io/fraud-tidbits/what-are-ad-fraud-solutions
16. What is Ad Fraud? Understanding, Detecting, and Preventing – ClickGUARD, дата последнего обращения: июня 4, 2025, https://www.clickguard.com/blog/what-is-ad-fraud/
17. What Are Click Bots? | Akamai, дата последнего обращения: июня 4, 2025, https://www.akamai.com/glossary/what-are-click-bots
18. Click Fraud Protection & Detection Software – Anura, дата последнего обращения: июня 4, 2025, https://www.anura.io/click-fraud
19. The Study on Preventing Click Fraud in Internet Advertising, дата последнего обращения: июня 4, 2025, https://csroc.cmex.org.tw/journal/JOC31-3/JOC3103-20.pdf
20. Research on Information Technology with Detecting the Fraudulent Clicks Using Classification Method – ResearchGate, дата последнего обращения: июня 4, 2025, https://www.researchgate.net/publication/284501084_Research_on_Information_Technology_with_Detecting_the_Fraudulent_Clicks_Using_Classification_Method
21. www.trafficguard.ai, дата последнего обращения: июня 4, 2025, https://www.trafficguard.ai/guides/why-machine-learning-is-the-future-of-ad-fraud-prevention#:~:text=Contextual%20Click%20Fraud%20Detection,between%20real%20users%20and%20bots.
22. Why Machine Learning is the Future of Ad Fraud Prevention – TrafficGuard, дата последнего обращения: июня 4, 2025, https://www.trafficguard.ai/guides/why-machine-learning-is-the-future-of-ad-fraud-prevention
23. Anomaly detection for fraud prevention – Advanced strategies, дата последнего обращения: июня 4, 2025, https://www.fraud.com/post/anomaly-detection
24. Anomaly detection in fraud prevention: Benefits, development and future trends, дата последнего обращения: июня 4, 2025, https://www.leewayhertz.com/anomaly-detection-in-fraud-prevention/
25. What are Behavioral Biometrics? | AuthenticID, дата последнего обращения: июня 4, 2025, https://www.authenticid.com/glossary/behavioral-analytics/
26. How Can Behavioral Analytics Help Detect Click Fraud? – Spider AF, дата последнего обращения: июня 4, 2025, https://spideraf.com/articles/how-can-behavioral-analytics-help-detect-click-fraud
27. www.ppchero.com, дата последнего обращения: июня 4, 2025, https://www.ppchero.com/the-ultimate-guide-to-click-fraud/#:~:text=Looking%20for%20Patterns%20in%20IP,by%20click%20farms%20and%20bots.
28. Browser Fingerprinting and Its Role in Fraud Detection – SHIELD, дата последнего обращения: июня 4, 2025, https://shield.com/blog/browser-fingerprinting-and-its-role-in-fraud-detection
29. Top 9 Browser Fingerprinting Techniques Explained – Bureau, дата последнего обращения: июня 4, 2025, https://www.bureau.id/blog/browser-fingerprinting-techniques
30. PPC Fraud Detection: 10 Strategies to Stop Fake Clicks and Save Your Ad Budget, дата последнего обращения: июня 4, 2025, https://clickpatrol.com/ppc-fraud-detection-10-strategies-to-stop-fake/
31. 5 Proven Strategies to Defeat Click Fraud in Your Ad Campaigns, дата последнего обращения: июня 4, 2025, https://fraudblocker.com/articles/click-fraud/5-proven-strategies-to-defeat-click-fraud-in-your-ad-campaigns
32. Expert Q&A: How to use honeypots to lure and trap bots – HUMAN Security, дата последнего обращения: июня 4, 2025, https://www.humansecurity.com/learn/blog/expert-q-a-how-to-use-honeypots-to-lure-and-trap-bots/
33. Why CAPTCHA Falls Short in Detecting Real Fraud | Anura Blog, дата последнего обращения: июня 4, 2025, https://www.anura.io/blog/why-captcha-is-ineffective-at-spotting-real-fraud
34. 8 Ways to Prevent Bots from Submitting Forms – Anura.io, дата последнего обращения: июня 4, 2025, https://www.anura.io/fraud-tidbits/ways-to-prevent-bots-from-submitting-forms
35. Advanced fraud detection – Stripe Documentation, дата последнего обращения: июня 4, 2025, https://docs.stripe.com/disputes/prevention/advanced-fraud-detection
36. Click Fraud Detection in Online Advertising: A Comparative Study of Machine Learning Models | IIETA, дата последнего обращения: июня 4, 2025, https://www.iieta.org/journals/ijsse/paper/10.18280/ijsse.150303
37. Advertisement Click Fraud Detection Using Machine Learning Algorithms – ResearchGate, дата последнего обращения: июня 4, 2025, https://www.researchgate.net/publication/360027156_Advertisement_Click_Fraud_Detection_Using_Machine_Learning_Algorithms
38. UNDERSTANDING AND MITIGATING AD FRAUD IN MENA, дата последнего обращения: июня 4, 2025, https://iabmena.com/understanding-and-mitigating-ad-fraud-in-mena/
39. Click Fraud Detection – Adsbot, дата последнего обращения: июня 4, 2025, https://adsbot.co/click-fraud-detection/
40. Leading Click Fraud Protection Software | Fraud Blocker™, дата последнего обращения: июня 4, 2025, https://fraudblocker.com/
41. Stop Click Fraud in Its Tracks: A PPC Guide – SocialSellinator, дата последнего обращения: июня 4, 2025, https://www.socialsellinator.com/social-selling-blog/ppc-click-fraud-prevention
42. Ad Click Defense: Advanced Click Fraud Protection for Platform, дата последнего обращения: июня 4, 2025, https://www.humansecurity.com/learn/blog/ad-click-defense-advanced-click-fraud-protection/
43. Best Practices For Click Fraud Prevention – ClickPatrol, дата последнего обращения: июня 4, 2025, https://clickpatrol.com/best-practices-for-click-fraud-prevention/
44. Google Click Fraud & How To Prevent It [2025 Guide] – PPC.io, дата последнего обращения: июня 4, 2025, https://ppc.io/blog/google-click-fraud
45. Facebook Ad Fraud Protection | TrafficGuard, дата последнего обращения: июня 4, 2025, https://www.trafficguard.ai/meta
46. Best Pay Per Click Fraud Detection Platform to Protect Your Ad Spend, дата последнего обращения: июня 4, 2025, https://www.anura.io/fraud-tidbits/best-pay-per-click-fraud-detection-platform
47. Top Ad-Fraud Suite Competitors & Alternatives 2025 | Gartner Peer Insights, дата последнего обращения: июня 4, 2025, https://www.gartner.com/reviews/market/online-fraud-detection/vendor/mfilterit/product/ad-fraud-suite/alternatives
48. Best Ad Verification Tools Reviews 2025 | Gartner Peer Insights, дата последнего обращения: июня 4, 2025, https://www.gartner.com/reviews/market/ad-verification-tools
49. Is Click Fraud Illegal? What the Law Says Across Different Countries, дата последнего обращения: июня 4, 2025, https://fraudblocker.com/articles/is-click-fraud-illegal
50. Truth In Advertising | Federal Trade Commission, дата последнего обращения: июня 4, 2025, https://www.ftc.gov/news-events/topics/truth-advertising
51. What Federal Crimes deal with Internet or Computer Fraud?, дата последнего обращения: июня 4, 2025, https://lgaulirufo.com/blog/exploring-federal-cyber-crimes-internet-and-computer-fraud/
52. Federal Computer Fraud (18 U.S.C. § 1030) – Leppard Law, дата последнего обращения: июня 4, 2025, https://leppardlaw.com/federal/fraud/federal-computer-fraud-18-u-s-c-%C2%A7-1030/
53. Cybercrime and the Law: Primer on the Computer Fraud and Abuse …, дата последнего обращения: июня 4, 2025, https://www.congress.gov/crs-product/R47557
54. What is PPC Click Fraud? – PPC Shield, дата последнего обращения: июня 4, 2025, https://www.ppcshield.io/knowledge-base/popular-questions/what-is-click-fraud/
55. Click Fraud Attorney, PPC Lawyer – Kronenberger Rosenfeld, LLP, дата последнего обращения: июня 4, 2025, https://kr.law/practice-areas/click-fraud
56. Нарушения в рекламе: штрафы и ответственность – Контур Эльба, дата последнего обращения: июня 4, 2025, https://e-kontur.ru/enquiry/2051/za-kakie-narusheniya-v-reklame-shtrafuet-fas
57. Understanding Click Fraud and How to Prevent It – SearchInform, дата последнего обращения: июня 4, 2025, https://searchinform.com/cybersecurity/cyber-threats/fraud/type/click-fraud/
58. How To Avoid Click Fraud In PPC – ClickPatrol, дата последнего обращения: июня 4, 2025, https://clickpatrol.com/how-to-avoid-click-fraud-in-ppc/
59. The 12 Most Common Types of Click Fraud – Fraud Blocker, дата последнего обращения: июня 4, 2025, https://fraudblocker.com/articles/click-fraud/types-of-click-fraud/the-12-most-common-types-of-click-fraud
60. Накрутка поведенческого фактора: все методы и их последствия – Skypro, дата последнего обращения: июня 4, 2025, https://sky.pro/wiki/marketing/nakrutka-povedencheskogo-faktora-vse-metody-i-ih-posledstviya/
61. Клик-фермы: что это такое и для чего нужны – Защита от скликивания рекламы, дата последнего обращения: июня 4, 2025, https://clickfraud.ru/klik-fermy-chto-eto-takoe-i-dlya-chego-nuzhny/
62. Click Fraud explained – PPC Land, дата последнего обращения: июня 4, 2025, https://ppc.land/click-fraud/
63. iabeurope.eu, дата последнего обращения: июня 4, 2025, https://iabeurope.eu/wp-content/uploads/IAB-Europe-Guide-to-Ad-Fraud-1-2.pdf
64. Скликивание в рекламе — что такое и как защититься от клик-фрода – СтопКлик, дата последнего обращения: июня 4, 2025, https://stopclick.ru/blog/clickfraud
65. KLiCK.kz – система накрутки кликов, скликивание рекламы, накрутка кликов, голосов, click-fraud, дата последнего обращения: июня 4, 2025, https://klick.kz/instruct.php?act=zakaz
66. Способы ручной накрутки поведенческих факторов – База знаний Timeweb Community, дата последнего обращения: июня 4, 2025, https://timeweb.com/ru/community/articles/sposoby-ruchnoy-nakrutki-povedencheskih-faktorov
67. Как конкуренты могут скликивать вашу рекламу и что с этим делать? – VC.ru, дата последнего обращения: июня 4, 2025, https://vc.ru/marketing/1788281-kak-konkurenty-mogut-sklikivat-vashu-reklamu-i-chto-s-etim-delat
68. AdTech: что нужно знать о рекламных технологиях в 2024 году, дата последнего обращения: июня 4, 2025, https://grandawards.ru/blog/reklamnye-tekhnologii-adtech/
69. What is Ad Fraud? | Understanding Ad Fraud | HUMAN Security, дата последнего обращения: июня 4, 2025, https://www.humansecurity.com/learn/topics/what-is-ad-fraud/
70. What is click fraud? | How click bots work | Cloudflare, дата последнего обращения: июня 4, 2025, https://www.cloudflare.com/learning/bots/what-is-click-fraud/
71. Клик-фермы: что это такое и для чего они нужны? – Защита от скликивания рекламы, дата последнего обращения: июня 4, 2025, https://clickfraud.ru/klik-fermy-chto-eto-takoe-i-dlya-chego-oni-nuzhny/
72. Это важно: как распознать и предотвратить рекламное мошенничество в онлайне, дата последнего обращения: июня 4, 2025, https://www.byyd.me/ru/blog/2020/10/ad-fraud/
73. What is Clickstream Data? Uses, Types & Dataset Examples – Datarade, дата последнего обращения: июня 4, 2025, https://datarade.ai/data-categories/clickstream-data
74. US8880441B1 – Click stream analysis for fraud detection – Google Patents, дата последнего обращения: июня 4, 2025, https://patents.google.com/patent/US8880441B1/en
75. Ad Click Fraud Detection Using Machine Learning and Deep Learning Algorithms, дата последнего обращения: июня 4, 2025, https://www.researchgate.net/publication/388213781_Ad_Click_Fraud_Detection_Using_Machine_Learning_and_Deep_Learning_Algorithms
76. www.iab.com, дата последнего обращения: июня 4, 2025, https://www.iab.com/wp-content/uploads/2015/05/IAB_Anti_Fraud_Principles_and_Taxonomy.pdf
77. Anomaly Detection in Python with Isolation Forest – DigitalOcean, дата последнего обращения: июня 4, 2025, https://www.digitalocean.com/community/tutorials/anomaly-detection-isolation-forest
78. Anomaly Detection in Machine Learning Using Python | The PyCharm Blog, дата последнего обращения: июня 4, 2025, https://blog.jetbrains.com/pycharm/2025/01/anomaly-detection-in-machine-learning/
79. Click Fraud Detection with Recurrent Neural Networks Optimized by Adapted Crayfish Optimization Algorithm – ACADlore, дата последнего обращения: июня 4, 2025, https://www.acadlore.com/article/JII/2024_2_4/jii020404
80. Новые штрафы за контекстную рекламу при использование чужих брендовых запросов, дата последнего обращения: июня 4, 2025, https://e-kontur.ru/enquiry/1467/kontekstnaya-reklama

Related publications:

Click fraud is a problem in today’s mobile advertising industry

How to Detect Click Fraud and Keep Your Advertising Budget from Being Wasted

In 2022, four common assumptions about advertising fraud will cost you money

What is click fraud and how does it happen? (part 1)

What is click fraud, and how does it affect you?

Click fraud statistics 2022

Making a stand against targeting bots: What is Click Fraud and How Can You Avoid It?

Click Fraud: Understanding the Danger