Chapter 1.2: The Gorgon Protocol: Adversarial Aesthetics and the Weaponization of the Grotesque

Volume 1: SCANNED — The Politics of Legibility

Chapter 1.2: Gorgon Aesthetics

1. Introduction: The Crisis of the Visible and the Biometric Imperative

The contemporary subject exists within a condition of "hyper-visibility." We are no longer merely seen; we are scanned, indexed, and correlated. The rise of Surveillance Capitalism, as articulated by Shoshana Zuboff, has transformed human experience into "behavioral surplus"—raw material for the prediction and modification of future behavior.¹ In this regime, the physical body is the primary interface of extraction. The face, once the locus of intersubjective empathy and communicative nuance, has been reconfigured by the mechanisms of the state and the market into a biometric key, a data anchor, and a marketable asset. To be "seen" by the machine is to be captured, categorized, and ultimately governed.

The "Ungovernable Body" trilogy posits a fundamental question: When the state and capital utilize the physical form as a data point, how does the revolutionary subject weaponize "un-readability"? This chapter, Gorgon Aesthetics, interrogates the intersection of ancient apotropaic magic, high-fashion grotesquerie, and adversarial machine learning to propose a radical counter-surveillance strategy: the weaponization of the "Hag."

We operate under the core hypothesis that high-contrast, asymmetrical, and "grotesque" makeup—styling that mimics the biological decay of aging, the chaotic noise of the glitch, or the "monstrous feminine"—acts as a functional adversarial attack against commercial Facial Recognition Technology (FRT). This approach diverges sharply from the geometric, often sleek aesthetic of "CV Dazzle," which creates "anti-faces" through cubist abstraction. Instead, Gorgon Aesthetics leverages the specific biases and failure modes of biometric algorithms regarding aging, texture, and gender. By simulating the "abject" features of the elderly female face—deep nasolabial folds, chaotic pigmentation, and structural asymmetry—the subject does not merely hide; they become "noise" to the system, exploiting the very algorithmic biases that render older women invisible in the public sphere.³

The ancient Greek Gorgoneion—the severed head of Medusa used as a protective amulet—functioned by turning the hostile gaze back upon itself.⁵ It was an image of such terrifying potency that it arrested the viewer, turning them to stone. Today, the hostile gaze is automated, embedded in the thousands of networked cameras that comprise the "smart city." To defeat this gaze, we must don the mask of the Gorgon. We must become the "Scary Old Woman," the entity that the algorithm, trained on the smooth, youthful faces of celebrity datasets, cannot process, cannot value, and therefore, cannot govern.

The imperative for this research is driven by the rapid enclosure of the "biological commons." As facial recognition becomes the standard for access control, banking, and policing, the right to not be recognized becomes the ultimate luxury good.⁶ However, traditional methods of evasion—masks, hoods—are increasingly criminalized or socially conspicuous. We require a form of camouflage that operates in plain sight, a "steganography of the skin" that allows the subject to move through the physical world unmolested while remaining opaque to the digital eye. This report provides the theoretical and technical framework for such a camouflage.

2. The Biometric Gaze: Architecture of the Digital Panopticon

To understand how the "grotesque" functions as camouflage, we must first dissect the visual logic of the machine. Facial Recognition Technology (FRT) does not "see" a face in the phenomenological sense. It processes topological data, searching for specific gradients of light and shadow that correspond to a statistical model of a human face. It is an act of mathematical translation, converting biological uniqueness into vector space.

2.1 The Mathematics of Face Detection: From Haar to Deep Learning

The fundamental architecture of modern FRT relies on detecting "landmarks." The industry standard, exemplified by libraries like Dlib, utilizes a 68-point annotation scheme (based on the iBUG 300-W dataset) to map the geography of the face.⁷ These points correspond to high-contrast regions: the corners of the mouth, the bridge of the nose, the curvature of the jaw, and the circumference of the eyes.

The evolution of this technology reveals its vulnerabilities:

1. Viola-Jones / Haar Cascades: This early method relies on simple rectangular features—for example, the observation that the eye region is generally darker than the cheek region. It scans the image for these contrasting rectangular blocks. While primitive, it is still used in low-power devices. Its weakness is a reliance on expected contrast.¹⁰

1. Histogram of Oriented Gradients (HOG): This method counts occurrences of gradient orientation in localized portions of an image. It relies heavily on edge detection and the predictable contrast between facial features and skin. It essentially draws a "sketch" of the face based on where light intensity changes.¹¹

1. Convolutional Neural Networks (CNNs) & Deep Learning: Modern systems (e.g., ResNet, VGG-Face) use deep learning to extract high-level semantic features. These models are trained on millions of images to learn what a face "is" at a conceptual level. However, they retain specific biases from their training data, particularly regarding texture and "face-ness".¹²

The vulnerability of all these systems lies in their reliance on predictable contrast and topology. A "normal" face has a smooth gradient across the cheek, a sharp gradient at the eyebrow, and a specific shadow pattern at the philtrum. When these gradients are disrupted—not just blocked, but corrupted with false data—the algorithm's confidence score plummets.

2.2 The "Bio-Age" Trap and Algorithmic Bias

Crucially, these algorithms are not neutral observers. They are historical artifacts, encoded with the biases of their creators and their training data. Extensive research, most notably by the National Institute of Standards and Technology (NIST) and the "Gender Shades" project by Joy Buolamwini, has demonstrated that FRT systems exhibit significant demographic disparities.³

The data reveals a stark hierarchy of visibility. The algorithm is optimized for the "Reference Man"—typically white, male, and middle-aged. As the subject deviates from this norm, the error rate increases.

Table 1: Algorithmic Error Rates by Demographic (NIST & MIT Studies)

Source: MIT News ¹⁴, CSIS ³, NIST Reports ¹⁵

The data indicates a "Pink Tax" of surveillance: women are harder to identify than men, and older subjects are harder to identify than younger ones.³ This is often framed by civil rights groups as a failure of inclusivity—a problem to be "fixed" by better data collection. However, for the resistance, this "failure" is a feature. It is a security vulnerability in the panopticon.

The algorithm's inability to accurately map the face of an older woman—specifically the texture of wrinkles, the loss of subcutaneous fat, and the shifting of landmarks due to gravity—creates a blind spot. The "Bio-Age" trap, where algorithms penalize signs of aging in other contexts (e.g., insurance pricing), here becomes a mechanism of escape. By artificially inducing the visual markers of extreme age and gendered "ugliness," the subject voluntarily enters the demographic category with the highest algorithmic failure rate. We do not ask for better representation in the database; we exploit our poor representation to escape it.

2.3 The Dlib 68-Point Paradigm and its Discontents

To understand the specific mechanism of the "Gorgon" attack, we must look at the Dlib 68-point landmark detector, the industry standard for facial alignment.⁸ Dlib locates:

• Points 1-17: Jawline

• Points 18-27: Eyebrows

• Points 28-36: Nose

• Points 37-48: Eyes

• Points 49-68: Mouth

The Dlib algorithm uses an ensemble of regression trees. It assumes that, statistically, pixel patch A (the eye) is usually at a certain vector distance from pixel patch B (the nose).

• The Semantic Gap: The algorithm does not "know" what an eye is. It knows that, statistically, a dark spot surrounded by lighter skin is an eye.

• The Gorgon Interruption: When we apply "Hag" makeup, we disrupt these vectors.

• Snippet ¹⁰ (Juggalo Makeup) highlights that blocking the jawline disrupts the entire mesh.

• Snippet ⁴¹ (Wrinkle Detection) notes that wrinkles are "curvilinear discontinuities."

• Synthesis: By drawing a fake wrinkle that intersects a landmark zone (e.g., a scar running through the eyebrow), we introduce a "discontinuity" that the regression tree cannot resolve. The tree "splits" the difference, resulting in a landmark placement that is wildly off-target. If the average landmark error (RMSE) exceeds a threshold, the face detection is discarded.

2.4 The Liveness Detection Failure: Texture as Defense

Modern FRT systems utilize "Liveness Detection" to prevent spoofing (e.g., holding up a photo or wearing a mask). These systems look for:

• Micro-movements: Blinking, breathing.

• Skin Texture: The scattering of light on human skin (subsurface scattering).

• Depth: 3D structure.

Hypothesis: Heavy, "grotesque" makeup, particularly prosthetics or liquid latex used to create wrinkles, creates a "mask" effect that can trigger a liveness detection failure.¹⁸

• Silicone/Latex Block: Snippet ¹⁹ confirms that custom silicone masks pose a serious threat to FRT. However, these are expensive. The "Gorgon" approach uses liquid latex and rigid collodion (scarring liquid) to create a similar effect cheaply.

• The "Uncanny" Signal: Snippet ⁵⁰ and ⁵¹ discuss the "uncanny valley." While this is usually a psychological effect, in computer vision, it manifests as a "low confidence" score. The system detects a face, but the texture maps don't match "living human skin"—they match "painted surface" or "dead tissue."

• Result: The user is flagged as a "spoof" or simply ignored. In a high-throughput surveillance scenario (e.g., a busy street), systems are tuned to minimize false positives. A "low confidence" face is often discarded to save processing power. To be identified as a "spoof" is to be ignored by the identification algorithm.

3. Historical Anthropology: The Apotropaic Gorgon

The strategy of "Gorgon Aesthetics" is not a novel invention of the digital age; it is a re-activation of ancient security protocols. The use of the monstrous face to repel harm is a technology as old as civilization itself, rooted in the anthropological concept of the apotropaic—magic designed to "turn away" evil.

3.1 The Gorgoneion as Security Protocol

In Ancient Greece, the image of the Gorgon (Medusa) was not merely a monster to be slain; it was a functional security device. The Gorgoneion—the disembodied head of Medusa, often depicted with a protruding tongue, boar tusks, and staring eyes—was placed on doors, shields, and city walls.⁵ This image was "apotropaic," meaning "to turn away." Its function was to terrify the viewer, to arrest their gaze, and to prevent them from crossing a threshold.

The Gorgoneion operates on a principle of "visual toxicity." It is an image that is dangerous to look at. In the myth, the gaze of Medusa turns the viewer to stone—an ultimate form of objectification. To look at the Gorgon is to cease to be a subject and become an object (a statue).

In the context of modern surveillance, the camera is the entity that turns us to stone. It captures the fluid, living subject and freezes them into a static data object, a digital "statue" stored in a server farm. The "Gorgon Aesthetic" attempts to reverse this polarity. By presenting the camera with an image that it cannot process—a "monstrous" data-object—we jam the mechanism of capture. The "Hag" face functions as a digital Gorgoneion: a shield that breaks the gaze of the scanner.

3.2 Baba Yaga and the Unseen Crone

Parallel to the Greek Gorgon is the Slavic figure of Baba Yaga. Described as a "terrifying crone" with iron teeth and a nose that touches the ceiling, Baba Yaga exists in the liminal space of the forest.²¹ She is an "ambiguous" figure, neither wholly good nor wholly evil, but defined by her power and her ungovernability.

Baba Yaga represents the "abject" woman—the post-reproductive female who has exited the economy of sexual desirability and reproductive labor. In patriarchal capitalism, this woman is invisible; she has no value. However, in the realm of surveillance, this invisibility is a superpower. The "Hag" is not tracked because she is not a target for advertising, nor is she easily categorized by systems trained on youthful, consumerist beauty standards.

Adopting the aesthetic of Baba Yaga—the "tattered rags," the "gnarled" features, the "wild hair"—is an act of "fleeing into the forest" of the dataset. It is a rejection of the "smoothness" required by the smart city.²¹ The "Hag" is the "outlier" that the system tries to smooth over or discard. By embodying the Hag, we become the statistical outlier that breaks the model.

3.3 The Evil Eye vs. The Electronic Eye

We can theorize the Surveillance Camera as the ultimate "Evil Eye" (Mal de Ojo). In Mediterranean and West Asian cultures, the Evil Eye is a curse cast by a malevolent gaze, often associated with envy. The surveillance camera is an "envious" eye—it seeks to extract value from the subject.

The traditional cure for the Evil Eye is often a reflection. The Nazar (blue eye amulet) or the Gorgoneion stares back.⁵ Wearing the Gorgon face is an act of reflective magic. It presents the camera with a "hyper-eye" (the exaggerated, painted eyes of the mask) that overwhelms its capacity to "curse" (identify) the subject. It is a visual feedback loop that destabilizes the observer.

4. The Aesthetic of Refusal: From CV Dazzle to Gorgon Glitch

The history of anti-surveillance aesthetics is brief but instructive. To understand why "Gorgon Aesthetics" represents a necessary evolution, we must analyze the limitations of previous attempts, most notably CV Dazzle.

4.1 The Limitations of CV Dazzle

Adam Harvey's CV Dazzle (Computer Vision Dazzle), developed around 2010, draws inspiration from "Dazzle Camouflage" used on WWI battleships—bold, geometric patterns designed to break up the silhouette of a ship so that enemies could not calculate its heading or speed.¹⁰ CV Dazzle works by creating "anti-faces," using cubist blocks of hair and high-contrast makeup to disrupt the symmetry that Haar cascade classifiers look for.

However, CV Dazzle has critical limitations in the 2020s:

1. High Visibility: It looks "cool." It borrows from cyberpunk and avant-garde fashion. While it confuses the machine, it attracts the human eye. In a physical security scenario, standing out is dangerous. It signals "I am hiding something."

1. Obsolescence: CV Dazzle was designed primarily for Haar cascades. Modern Deep Convolutional Neural Networks (DCNNs) are much more robust. They are trained on millions of "in-the-wild" images, including occluded faces. They can infer a face even if parts are missing.¹⁰ The geometric abstraction of CV Dazzle is increasingly readable to advanced AI.

1. Fashion Capture: The aesthetic of CV Dazzle has been recuperated by the fashion industry. It reads as "edgy," not "glitchy." It has been commodified.

4.2 The "Juggalo" Anomaly: Accidental Adversarialism

A fascinating and accidental discovery in adversarial aesthetics is the efficacy of Juggalo makeup. The insignia of the "Insane Clown Posse" fanbase involves heavy black and white paint, often covering the chin and mouth in a grimace.

Research indicates that Juggalo makeup is surprisingly effective against FRT.¹⁰ This is because it completely obliterates the jawline and mouth landmarks—two critical anchor points for Dlib and other shape predictors. The high-contrast black paint on the jaw reshapes the lower face, causing the algorithm to miscalculate the "bounding box" of the face.

However, like CV Dazzle, Juggalo makeup is culturally loaded and highly visible. It signals membership in a specific subculture, which is itself a form of data classification. It allows for "tribal" identification even if "individual" identification fails.

4.3 Enter the Gorgon: The "Glitch" Aesthetic

"Gorgon Aesthetics" proposes a synthesis of the Juggalo's high-contrast masking and the "Hag's" organic irregularity. It moves away from the geometric (CV Dazzle) toward the organic, the chaotic, and the visceral.

This aligns with Glitch Feminism, as theorized by Legacy Russell. Russell argues that the "glitch" is not an error, but a refusal to function within the binary logic of the machine.²⁷ The glitch is "opacity." It is the refusal to be readable.

Applying this to makeup means embracing:

• Asymmetry: The human face is naturally symmetrical. The algorithm expects this. The Gorgon face is radically asymmetrical (e.g., one eye heavily lined, the other bare; a "scar" running diagonally across the face).

• Texture Overload: Simulating skin diseases, burns, or extreme wrinkling adds high-frequency noise to the image. This confuses texture-based liveness detection and age-estimation algorithms.¹²

• The Grotesque: Utilizing the aesthetics of Rick Owens' "vicious" runway shows—where models appear as "aliens" or "witches"—to create a visual persona that triggers an "uncanny valley" response in both humans and machines.³⁰

Table 2: Comparative Efficacy of Privacy Aesthetics

5. Fashion Theory and the Grotesque: Rick Owens as Case Study

The "Gorgon" is not just a data-mask; it is a performance. We turn to fashion theory to understand how "monstrous" aesthetics function in the public sphere, specifically looking at the work of Rick Owens and Rei Kawakubo (Comme des Garçons).

5.1 Rick Owens: The Autobiography of Damage

Rick Owens' fashion philosophy provides the perfect aesthetic template for the Ungovernable Body. He describes his work as "an expression of tenderness and raging ego... the damage I've done on the way".³² This aligns with the "Hag" aesthetic—the face as a record of damage (time, sun, grief).

In his SS14 "Vicious" show, Owens used step-dancers with "grit-teeth" expressions and aggressive posturing.³⁰ This performance rejected the passive, blank stare of the traditional model. The models were threatening. In the context of surveillance, the passive face is the readable face. The "grit-teeth," screaming, distorted face (the "Gorgon" face) disrupts the "neutral expression" assumption of passport photos and ID scans. To walk through the city with a "vicious" face is to break the social contract of surveillance, which demands docility.

Owens' aesthetic is one of "exquisite loneliness" and "sovereignty".³³ To dress in a way that repels the gaze is to assert ownership over one’s own image. In the context of the smart city, this is a radical act. It says: "I am not here for your consumption. I am here for my own purpose, and that purpose is ungovernable."

5.2 Comme des Garçons: The Body as Lump

Rei Kawakubo of Comme des Garçons often distorts the human silhouette with lumps, bumps, and asymmetries, most famously in the "Body Meets Dress, Dress Meets Body" collection (often called the "Lumps and Bumps" collection).

This aesthetic has direct relevance to Gait Recognition. Modern surveillance uses gait analysis to identify subjects by their walk cycle and silhouette.³⁴ Research snippet ⁵² notes that altering the silhouette (e.g., wearing a skirt vs. pants, or carrying a heavy bag) can affect gait recognition accuracy.

The "Gorgon" aesthetic, with its voluminous, layered, "Rick Owens-esque" drapery ³⁵ and the simulated "stoop" of the Hag, completely obscures the skeletal kinematics required for gait analysis. By creating a "lumpy" silhouette, the subject prevents the algorithm from identifying the joints (knees, hips) necessary to model the gait. The "Body as Lump" is an encrypted body.

5.3 "Hag Drag" and the Monstrous Feminine

The "Hag" has seen a resurgence in horror media and "Hag Drag".³⁶ Drag queens have long used "clown white" and extreme contouring to reshape the face. "Hag Drag" (mimicking the Witch from Snow White, etc.) ³⁸ is a sub-genre that celebrates the power of the "evil" woman.

We can learn from Drag techniques. The use of "blocking" eyebrows (gluing them down and painting new ones higher up) is a standard Drag technique. It is also a highly effective anti-FRT technique because it moves the "eyebrow" landmark to the forehead.³⁹ The "Gorgon" protocol adopts these theatrical techniques but applies them for survival rather than entertainment.

6. Technical Implementation: The Gorgon Aesthetic Protocol

Based on the synthesis of fashion theory, security research, and makeup artistry, we can codify the "Gorgon Aesthetic" into a practical protocol for adversarial defense.

6.1 Protocol G-1: The "Medusa" Glitch (Landmark Disruption)

Target: Dlib 68-Point Landmarker (Jaw, Mouth, Nose)

Materials: Rigid Collodion, Spirit Gum, Black/Grey Grease Paint.

1. Jawline Dissolution: Apply patches of spirit gum and cotton along the jawline to create irregular "boils" or "scars." Paint over with a foundation 3 shades lighter than skin tone. This destroys the strong gradient edge of the jaw, confusing the HOG detectors.¹⁰

1. Nasolabial Deepening: Apply 5-7 layers of Rigid Collodion to the nasolabial folds. As it dries, it puckers the skin, creating a deep, physical trench. Shade the inside with dark grey. This creates a "shadow trap."

• Adversarial Effect: Under overhead surveillance lighting, these artificial trenches create hard black lines. The Haar cascade or HOG detector sees a "mouth" that is three times wider than it should be, or a "nose" that connects to the chin. The landmark topology is shattered.⁴¹

1. The "Evil Eye" Brow: Glue down the natural eyebrows. Paint new, asymmetrical eyebrows 1 inch higher on the forehead. Use a jagged, "hairy" stroke pattern. This confuses the "eye-to-eyebrow" distance metric, a key biometric identifier.³⁹

6.2 Protocol G-2: The "Texture" Attack (Liveness Failure)

Target: Liveness Detection (Subsurface Scattering, Texture Analysis)

Materials: Liquid Latex, Tissue Paper, Coarse Sponge.

1. Latex Stippling: Apply liquid latex over the cheeks and forehead using a coarse sponge. While wet, press crumpled tissue paper into it and peel off. This creates a "micro-cratered" texture.

1. Color Noise: Stipple varying shades of red, purple, and yellow over the latex to simulate "liver spots," broken capillaries, or skin disease.

1. Adversarial Effect: The skin no longer reflects light uniformly. The "specular highlights" (shiny spots) are randomized. This disrupts "Shape-from-Shading" algorithms and triggers "spoof" alerts in Liveness detectors.¹⁹ The texture analysis sees "high entropy" (chaos) instead of the smooth entropy of skin, leading to a rejection of the face as a "mask" or "object."

6.3 Protocol G-3: The "OOD" (Out of Distribution) Palette

Target: CNN Classifiers (Demographic Grouping)

Materials: High-pigment primary colors (Yellow, Blue, Green).

1. Undertone Shift: Mix a strong yellow or blue pigment into the foundation. The goal is to create a skin tone that does not exist in the training set (which is usually reddish/brown/beige).

1. Adversarial Effect: Machine learning models are fragile when presented with data that lies outside their training distribution (Out-of-Distribution or OOD).⁴³ An image of a woman with "Gorgon" styling—extremely pale or dark skin with unnatural undertones, chaotic pigmentation, and exaggerated wrinkles—is a statistical outlier.

1. Energy Score Spike: OOD detection often uses "energy scores." A Gorgon face, with its high-frequency noise (wrinkles/makeup) and low semantic familiarity, yields a high energy score, causing the model to classify it as "noise" or "non-face".⁴⁵

7. Sociopolitical Synthesis: The Right to Rot

The adoption of the "Gorgon Aesthetic" is not merely a technical hack; it is a political stance. It challenges the "Biometric Imperative"—the societal demand that we present ourselves as "good data."

7.1 Reclaiming the Abject

Julia Kristeva defines the "abject" as that which "disturbs identity, system, order." The corpse, the open wound, the decaying body. In a culture obsessed with hygiene, youth, and high-definition visibility, the "aging" female body is abject.

By voluntarily becoming abject—by performing "hagdom"—the subject exits the marketplace. The surveillance capitalist economy wants to predict your behavior to sell you "anti-aging" cream.⁴⁶ It does not know what to sell to a witch. The "Hag" is a "bad consumer" and therefore a "low-value target" for data extraction. The refusal to optimize one's appearance is a refusal to participate in the "aesthetic labor" demanded by the service economy.⁴⁸

7.2 The "Pink Tax" of Privacy

We must return to the irony of the NIST data: The system fails most often on older women.³

• Thesis: This failure is a form of privilege that has been framed as a deficit.

• Action: We must stop asking for "better" algorithms that can recognize our grandmothers. We must instead universalize the grandmother's opacity.

• The "Gorgon" aesthetic democratizes this opacity. It allows a 20-year-old activist to don the "digital armor" of a 70-year-old crone. It redistributes the "Pink Tax" of surveillance failure, turning a bug into a shield.

7.3 Legal Implications: The "Makeup" Loophole

Many jurisdictions have anti-mask laws (e.g., prohibiting covering the face at protests). However, makeup occupies a legal grey area.

• The "Enhancement" Defense: Cosmetic makeup is legally protected as a form of expression or grooming. It is difficult to legislate against "bad makeup." If a subject claims their "Gorgon" makeup is simply "avant-garde fashion" or "poorly applied contouring," it is harder to prosecute than wearing a balaclava.

• Plausible Deniability: The Gorgon aesthetic, unlike the Guy Fawkes mask, relies on plausible deniability. It mimics a natural (if extreme) biological state (aging/illness). It forces the state to adjudicate "ugliness," a subjective category that the law is ill-equipped to handle.

8. Conclusion: Cryptographic Camouflage

The "Scary Old Woman" is not a failed aesthetic; it is a superior security protocol.

In the "Ungovernable Body" trilogy, we seek strategies that do not merely hide the subject but redefine the subject's relationship to power. Standard encryption protects data after it has been generated. Gorgon Aesthetics protects the subject before they become data.

By simulating the "noise" of biological decay—wrinkles, asymmetry, chaos—we exploit the structural weaknesses of an artificial intelligence trained on the smooth, symmetrical lies of the beauty industry. We weaponize the "uncanny valley." We invoke the ancient power of the Gorgon to turn the digital eye to stone.

Ugliness, in this framework, is not an absence of beauty. It is an excess of information. It is cryptographic camouflage. It is the only way to walk through the digital panopticon and remain, gloriously, unread.

The "Hag" is the ghost in the machine. She is the outlier that cannot be smoothed. She is the error that persists. In the age of total surveillance, the only true freedom is the freedom to be monstrous.

(End of Report)

Works cited

1. 'Surveillance Capitalism, COVID-19 and Social Work': A Note on Uncertain Future(s) - PMC, accessed on December 19, 2025, <u>https://pmc.ncbi.nlm.nih.gov/articles/PMC8344756/</u>

1. The Age of Surveillance Capitalism by Shoshana Zuboff review – we are the pawns \| Society books \| The Guardian, accessed on December 19, 2025, <u>https://www.theguardian.com/books/2019/feb/02/age-of-surveillance-capitalism-shoshana-zuboff-review</u>

1. The Problem of Bias in Facial Recognition \| Strategic Technologies Blog - CSIS, accessed on December 19, 2025, <u>https://www.csis.org/blogs/strategic-technologies-blog/problem-bias-facial-recognition</u>

1. Biased Technology: The Automated Discrimination of Facial Recognition \| ACLU of Minnesota, accessed on December 19, 2025, <u>https://www.aclu-mn.org/en/news/biased-technology-automated-discrimination-facial-recognition</u>

1. The eyes have it - The University of Sydney, accessed on December 19, 2025, <u>https://www.sydney.edu.au/museum/news/2022/10/04/the-eyes-have-it.html</u>

1. Identification of facial landmarks using Dlib. a Facial landmarks. b The position and order of 68 points on the face - ResearchGate, accessed on December 19, 2025, <u>https://www.researchgate.net/figure/dentification-of-facial-landmarks-using-Dlib-a-Facial-landmarks-b-The-position-and_fig2_343699139</u>

1. Web application for facial wrinkle recognition - CEUR-WS.org, accessed on December 19, 2025, <u>https://ceur-ws.org/Vol-3077/paper19.pdf</u>

1. Facial mapping (landmarks) with Dlib + python \| by Italo José \| TDS Archive - Medium, accessed on December 19, 2025, <u>https://medium.com/data-science/facial-mapping-landmarks-with-dlib-python-160abcf7d672</u>

1. Guide to detect Facial features using dlib - Kaggle, accessed on December 19, 2025, <u>https://www.kaggle.com/code/vishakha10/guide-to-detect-facial-features-using-dlib</u>

1. Subtle makeup tweaks can outsmart facial recognition - The Register, accessed on December 19, 2025, <u>https://www.theregister.com/2025/01/15/make_up_thwart_facial_recognition/</u>

1. Practical-CV/Facial-Landmarks-Detection-with-DLIB - GitHub, accessed on December 19, 2025, <u>https://github.com/Practical-CV/Facial-Landmarks-Detection-with-DLIB</u>

1. Classifying Conditions of Speckle and Wrinkle on the Human Face: A Deep Learning Approach - MDPI, accessed on December 19, 2025, <u>https://www.mdpi.com/2079-9292/11/21/3623</u>

1. Fashion-Guided Adversarial Attack on Person Segmentation - CVF Open Access, accessed on December 19, 2025, <u>https://openaccess.thecvf.com/content/CVPR2021W/WMF/papers/Treu_Fashion-Guided_Adversarial_Attack_on_Person_Segmentation_CVPRW_2021_paper.pdf</u>

1. Study finds gender and skin-type bias in commercial artificial-intelligence systems, accessed on December 19, 2025, <u>https://news.mit.edu/2018/study-finds-gender-skin-type-bias-artificial-intelligence-systems-0212</u>

1. The Critics Were Wrong: NIST Data Shows the Best Facial Recognition Algorithms Are Neither Racist Nor Sexist \| ITIF, accessed on December 19, 2025, <u>https://itif.org/publications/2020/01/27/critics-were-wrong-nist-data-shows-best-facial-recognition-algorithms/</u>

1. Artificial Intelligence Confirming Treatment Success: The Role of Gender- and Age-Specific Scales in Performance Evaluation - PMC - NIH, accessed on December 19, 2025, <u>https://pmc.ncbi.nlm.nih.gov/articles/PMC9512241/</u>

1. Facial landmarks with dlib, OpenCV, and Python - PyImageSearch, accessed on December 19, 2025, <u>https://pyimagesearch.com/2017/04/03/facial-landmarks-dlib-opencv-python/</u>

1. \[2205.13412\] A Physical-World Adversarial Attack Against 3D Face Recognition - arXiv, accessed on December 19, 2025, <u>https://arxiv.org/abs/2205.13412</u>

1. Spoofing Deep Face Recognition with Custom Silicone Masks - Idiap Publications, accessed on December 19, 2025, <u>https://publications.idiap.ch/downloads/papers/2018/Bhattacharjee_BTAS2018_2018.pdf</u>

1. The Dread Gorgon \| Lapham's Quarterly, accessed on December 19, 2025, <u>https://www.laphamsquarterly.org/fear/dread-gorgon</u>

1. Baba Yaga - The Paganista, accessed on December 19, 2025, <u>https://www.thepaganista.com/blog/baba-yaga</u>

1. Riding with Baba Yaga - Dangerous Women Project, accessed on December 19, 2025, <u>https://dangerouswomenproject.org/2017/01/11/riding-baba-yaga/</u>

1. The Unapology of Baba Yaga - House of Gamut, accessed on December 19, 2025, <u>https://www.houseofgamut.com/magazine/the-unapology-of-baba-yaga/</u>

1. Bizarre Makeup Patterns Can Fool Face Recognition Software - Discover Magazine, accessed on December 19, 2025, <u>https://www.discovermagazine.com/bizarre-makeup-patterns-can-fool-face-recognition-software-29104</u>

1. Adv-Makeup: A New Imperceptible and Transferable Attack on Face Recognition - IJCAI, accessed on December 19, 2025, <u>https://www.ijcai.org/proceedings/2021/0173.pdf</u>

1. It turns out that Juggalo makeup blocks facial recognition technology - California Lawyers Association, accessed on December 19, 2025, <u>https://calawyers.org/law-practice-management-technology/it-turns-out-that-juggalo-makeup-blocks-facial-recognition-technology/</u>

1. Glitch Feminism: A Manifesto \| Verso Books, accessed on December 19, 2025, <u>https://www.versobooks.com/products/460-glitch-feminism</u>

1. The Multiple Possibilities of Error: \[Glitch Feminism\] - terremoto.mx, accessed on December 19, 2025, <u>https://terremoto.mx/revista/56064/the-multiple-possibilities-of-error-glitch-feminism?lang=en</u>

1. Evaluation of Automatic Facial Wrinkle Detection Algorithms - PMC - PubMed Central, accessed on December 19, 2025, <u>https://pmc.ncbi.nlm.nih.gov/articles/PMC8321017/</u>

1. Is this the beginning of a revolution? Womenswear - Dazed, accessed on December 19, 2025, <u>https://www.dazeddigital.com/fashion/article/17373/1/is-this-the-beginning-of-a-revolution</u>

1. Paris Fashion Week: When Rick Owens' grotesque models and alien styles frightened the crowd \| South China Morning Post, accessed on December 19, 2025, <u>https://www.scmp.com/magazines/style/fashion-beauty/article/2188214/paris-fashion-week-when-rick-owens-grotesque-models</u>

1. Rick Owens' Aesthetic Vocabulary: Decoding the Designer's Vision \| Polimoda Fashion School, accessed on December 19, 2025, <u>https://www.polimoda.com/rick-owens-aesthetic-vocabulary-fashion-philosophy/</u>

1. Rick Owens' Exquisite Loneliness - Liberties Journal, accessed on December 19, 2025, <u>https://libertiesjournal.com/online-articles/rick-owens-exquisite-loneliness/</u>

1. Bilevel Optimized Collusion Attacks Against Gait Recognizer \| IEEE Journals & Magazine, accessed on December 19, 2025, <u>https://ieeexplore.ieee.org/document/10793411/</u>

1. Avant-Apocalypse: Do Androids Dream of Rick Owens? - Business & Arts, accessed on December 19, 2025, <u>https://businessandarts.net/blog/do-androids-dream-of-rick-owens</u>

1. The Hackneyed Hag - Blood Knife, accessed on December 19, 2025, <u>https://bloodknife.com/the-hackneyed-hag/</u>

1. Hag Makeup Demo - Snow White Witch - YouTube, accessed on December 19, 2025, <u>https://www.youtube.com/watch?v=Airf5x2584s</u>

1. The Old Hag (from Snow White) Makeup Tutorial - Disney Villains 3/9 - YouTube, accessed on December 19, 2025, <u>https://www.youtube.com/watch?v=E7-sFl3SHU8</u>

1. Yes, Your Makeup IS Aging You! \| 8 Mistakes You Might Be Making & How To Fix Them, accessed on December 19, 2025, <u>https://www.youtube.com/watch?v=4NLqoka68i0</u>

1. Makeup Changes the Apparent Size of Facial Features - The Cupola: Scholarship at Gettysburg College, accessed on December 19, 2025, <u>https://cupola.gettysburg.edu/cgi/viewcontent.cgi?article=1087&context=psyfac</u>

1. Edge‐based effective active appearance model for real‐time wrinkle detection - NIH, accessed on December 19, 2025, <u>https://pmc.ncbi.nlm.nih.gov/articles/PMC8247305/</u>

1. Makeup Presentation Attacks: The Deceptive Art of Fooling Facial Recognition, accessed on December 19, 2025, <u>https://antispoofing.org/makeup-presentation-attacks-techniques-attack-instruments-and-countermeasures/</u>

1. Accurately Classifying Out-Of-Distribution Data in Facial Recognition - Emory University, accessed on December 19, 2025, <u>http://www.math.emory.edu/site/cmds-reuret/projects/2023-ood/content/2023-OOD-Poster.pdf</u>

1. \[2404.03876\] Accurately Classifying Out-Of-Distribution Data in Facial Recognition - arXiv, accessed on December 19, 2025, <u>https://arxiv.org/abs/2404.03876</u>

1. Out-of-Distribution Detection with Adversarial Outlier Exposure - CVF Open Access, accessed on December 19, 2025, <u>https://openaccess.thecvf.com/content/CVPR2025W/SAIAD/papers/Botschen_Out-of-Distribution_Detection_with_Adversarial_Outlier_Exposure_CVPRW_2025_paper.pdf</u>

1. Late Capitalism, Beauty Standards and the War Against Aging - Kenn Orfanos, accessed on December 19, 2025, <u>https://kennorphan.com/2025/11/28/late-capitalism-beauty-standards-and-the-war-against-aging/</u>

1. Biometric Face Masks: What They Are and How They Benefit Your Skin \| Authentic K-Beauty, accessed on December 19, 2025, <u>https://koreancosmetics.cy/biometric-face-masks-what-they-are-and-how-they-benefit-your-skin/</u>

1. Surveillance and self-care: Performing beauty labor in online video content - Erasmus University Rotterdam, accessed on December 19, 2025, <u>https://pure.eur.nl/en/publications/surveillance-and-self-care-performing-beauty-labor-in-online-vide/</u>

1. Full article: Aesthetic citizenship, beauty politics and the state: an introduction, accessed on December 19, 2025, <u>https://www.tandfonline.com/doi/full/10.1080/13621025.2024.2363623</u>

1. Neural correlates of the uncanny valley effect for robots and hyper-realistic masks - PMC, accessed on December 19, 2025, <u>https://pmc.ncbi.nlm.nih.gov/articles/PMC11957730/</u>

1. Familiarity, orientation, and realism increase face uncanniness by sensitizing to facial distortions \| JOV \| ARVO Journals, accessed on December 19, 2025, <u>https://jov.arvojournals.org/article.aspx?articleid=2778703</u>

1. \[2002.09674\] Temporal Sparse Adversarial Attack on Sequence-based Gait Recognition, accessed on December 19, 2025, <u>https://arxiv.org/abs/2002.09674</u>