Face Similarity Linkage

A Novel Biometric approach to sexually motivated serial killer victims – Face Similarity Linkage 

This study focus on the sexual motivation of serial killers. There are cases without any Forensic evidence available from the crime scene. The appearances of victims could be a factor that the serial killers consider before committing a crime. Biometric approaches can be used as crime linkage in the absence of evidence at the crime scene. Physical evidence such as DNA can link an offender to the crime. Even without the evidence, the offender can be linked to the crime by analyzing the decisions and behavior of the offender and is known as behavioral fingerprinting. Crime linkage depends upon two main assumptions – a) consistency in the behavior of the offender,b) distinctiveness or variations in the behavior of the offender.

Typology of serial killers

Researchers divide serial killers into two different categories- Organised and disorganized offenders. Organized offenders are very intelligent, highly organized, planned, socially adequate, sexually competent, live with their partners, and have very high IQ.  Organized offender’s victim’s selection is based on appearances and is organized. Disorganized offenders select victims of their opportunity near their home or workplace. They are situational victims. Offenders usually chose victims based on the victim’s availability,  desirability (desire and sexual attraction of offenders to the victim ), and vulnerability (circumstances in the victim’s life to victimize them).


Biometrics differentiates one person from another based on body parts, biological, chemical characteristics, and behavior. Biometric systems are in the commercial, government, and forensic areas.

Facial Biometrics

Facial biometrics is the comparison of two or more faces for similarities and dissimilarities. In this study, faces are compared for similarities. According to Facial Identification scientific working group guidelines (FISWG), there are four different methods of facial comparison – Holistic comparison, Morphological Analysis, and Photo-anthropometry.

Holistic Comparison

Holistic comparison is a facial review process of an image with image sets for comparison. The process is used in Forensic Analysis. This process does not require any specialized equipment or training, and the time required for analysis is less. This procedure’s disadvantage is there is no precise documentation, and comparison accuracy rates are low, variable.

Morphological Analysis

Morphological analysis is the shape, appearance, and location of facial features. The morphological analysis does not require high-resolution images.


This process entails overlaying two images and comparing them visually. The advantage is that it enhances morphological analysis. The disadvantage is that it requires high image quality and viewpoint and the same for each image for comparison purposes.

Facial photo anthropometry

A facial photo anthropometry is a metric approach to two-dimensional images by comparing the images to reference images on a database known as Forensic Facial Identification.

Methods-Facial Comparisons and measurements

The research method employed is a simple and hand-measured extension of Forensic Facial Identification i.e., Face Similarity Linkage. This process is not an expert system method and consist of manually collected measurement. Facial measurement data of three different victims of a single serial killer have been compiled together to explore similarities and dissimilarities. Comparison of 3 randomly selected victims of the serial killer (Ted Bundy) and a randomly selected female reference image was collected and measured.

Identification and location of anatomical Landmarks

Twenty-seven landmarks were identified with definitions, abbreviations sourced from Lee et al. (2019) and Caple, Stephan (2016). These landmarks are presented in Table 1.



Measurement of Landmarks

Images are publicly accessible from an online source. Landmarks are chosen because of their ease of identification across all images by analysts and laypersons. Landmarks are mapped to 4 different subjects. Subjects included three victims (A, B, C) of a single serial killer and a female image (D) from the University of Yale database. Female image (D) is considered as an outgroup. Images are printed onto A3 paper.

To measure facial traits, a caliper was used with a measurement accuracy of 0.01mm. Images do not have an identifiable scale so that proportionality indices were measured. Measurement details are in Table 2. The numbers obtained are multiplied by 100 to give the proportion of that feature. Regarding angular measurements, the horizontal lines in figure 1 were used as a baseline to determine α for zygomatic (cheekbone) arch and eyebrow arch. Differences in the frontals of the image (bone related to forehead) from right and left were averaged.

Face Similarity Linkage


Statistical Analysis

Z scores were calculated for each facial trait. Z scores are used to determine P-value. A p-value of 0.5 was used to determine statistical significance (if the p-value is  ≥ 0.5, then the difference is not significant, and the p-value is ≤ 0.5, then the difference is significant).


The mean and standard deviation of all facial traits were calculated for each image and compared to the reference image (D). Statistical Analysis allowed comparison of all subjects and is represented in Table 3. The standard deviation of the images (A, B, C) was low and showed less deviation from the mean of the trait. P values and z scores of images A, B, and C were compared, and there were no statistical differences (p≥0.5). Comparison of subjects A, B, C to D showed p values traits that were statistically significant. Table 1 showed subject D has statistical significance when compared to A, B, and C. Image D is geometrically more different in facial structure than A, B, and C.  Figure 2 shows the comparison of Ted Bundy’s victim’s proportionality indices to the indices of image D.


Face Similarity Linkage



Face Similarity Linkage



The current study is a novel method for crime linkage based on the victim’s facial biometrics. Douglas et al., in 2016, stated that serial killers get attracted to the victims based on common characteristics such as occupation, gender, appearances, and social status. This research study focuses on the similarity in the appearance of victims and measures them in a scientific approach. The serial killer, Ted Bundy’s victims, exhibit similarities without any statistical significance in their proportionalities but the subjects (A, B, C) differ from the female reference image (B) with statistical significance.

Applications of Face Similarity Linkage

Facial anthropometry can be used in Forensic Analysis when all other tools are exhausted and considered as an extra tool for investigation. This procedure is not a solution for every serial crime. It should be noted that if there is a significant landmark in one of the victims than others, it doesn’t indicate that the person is not linked to the single offender. Forensic Facial Identification is an effortless and inexpensive procedure that can be processed in very little time. There is no need for any equipment and extensive training. This tool is not a complete solution. The time and geography of potential victims should be considered. This method should be progressed from manual to an automated software procedure.

Limitations of Face Similarity Linkage

The use of facial biometrics can make the procedure unreliable because of the changes in illumination, facial expression, occlusion, etc. Caution should be taken while using facial biometric techniques. Techniques should be improved in the future from manual to more sophisticated methods.

Leave a Comment