The Final Signature

Unless mail-in voting will be banned, the US democracy is in imminent threat from AI, one which will allow a government to control the results of the upcoming 2024 United States presidential election.

The Signature Mess

As I’ve written in the past, security is a social construct. Without free and fair elections there can be no democracy. In a democratic society, the election system is the most important information system that exists (in society), because without it there can be no trust between citizens, and without trust there can be no democracy.

In the United States, the process of casting a vote through the mail, also known as postal voting, mail-in voting, or vote by mail, is a type of absentee ballot. In 2020, as a result of COVID19, mail-in voting has increased drastically. According to data from the United States Election Assistance Commission (EAC), approximately 65.8 million Americans voted by mail or absentee ballot in the 2020 elections. Since there is a high probability that this trend might continue or even increase, it is essential for us to first take a look at the current state of affairs when it comes to the situation as it is right now, before we delve into the dangers from AI.

Mail-in voting involves sending a ballot directly to the registered voter's residence, allowing them to complete it and return it either through postal mail or by personally depositing it in a secure dropbox or voting center. To prevent forged paper ballots, most states employ signature checks. While some states allow in-person voting without ID requirements, most have verification procedures for mail-in ballots. In 18 states, a signature on the ballot envelope suffices without further verification. However, the remaining 32 states verify if the signature matches one on file, and some even necessitate witnesses or notaries. Upon receiving mailed ballots, the initial step involves comparing the voter's signature on the envelope with signatures stored in the election office. Smaller jurisdictions typically assign temporary staff to perform this task, while larger ones employ computers to scan the envelopes and swiftly determine if the signatures match.

When it comes to manual verification, handwriting experts assert that determining the authenticity of a signature or limited writing sample is extremely challenging. When manual signature reviews occur, election officials with limited or no training in signature verification are tasked with this responsibility. It is already unlikely that only one or two samples can capture the full range of a person's normal signature variations, making the verification process even more complex.

CISA's "Signature Verification and Cure Process" details how you verify signatures.
Mail ballot sorting equipment and voter registration system use Automated Signature Verification (ASV) applications that use a camera to capture the voter’s signature from the ballot return envelope & compared with reference image from voter registration database. According to the CISA guide, the usage of signature verification software is considered "Tier 1". It means that if the software provided a "correct" answer, the ballot is not required to pass any further verification.

Currently, non-matching signatures are required to be segregated into a separate bin, and temporary staff then conduct a double-check on the rejections, and in some locations, they also review the accepted envelopes.

Error Rate

The error rates of computerized signature reviews are not publicly disclosed, although academic researchers have reported error rates ranging from 10% to 14%. Various algorithms and standards, specific to the machine manufacturer, are employed to verify signatures. Furthermore, counties possess the discretion to manage settings and implement manufacturers' guidelines. Unfortunately, According to Stanford Law School report from 2020, entitled “Signature Verification and Mail Ballots: Guaranteeing Access While Preserving Integrity”, there are no standardized statewide criteria for automatic signature verification, and most counties do not provide publicly available, written explanations of their signature verification processes.

The mail-in voting system is as safe and effective as much as Pfizer/BioNTech mRNA “vaccines” are. You can read about the mess of mail-in voting as it presents itself in different US states in the WikiPedia page on the topic.

What Is a Signature?

Let us start with the basics - what is a signature? In "Modeling the Lexical Morphology of Western Handwritten Signatures", Diaz-Cabrera et al focused on the lexical morphology of handwritten signatures, meaning the identification, analysis, and description of the signature structures.

Modeling the Lexical Morphology of Western Handwritten Signatures.

According to the paper, Western signatures are generally composed of text (reflects the name, the family name etc) plus a flourish, or rubric in the occidental signatures which is defined by a kind of doodle written much faster and without much attention. The researchers used a small set of possible parameters to gain a better understanding of the main factors which characterize the way the signatures are performed. But here are multiple parameters which construct a signature:

• Form: The form of a handwritten signature refers to how you write the letters of your name. It can be in cursive (joined letters) or print (separate letters). Your signature's form adds uniqueness and individuality to your identification, making it distinguishable from others.

• Line quality: Line quality describes the thickness, smoothness, and consistency of the pen or pencil strokes used in your signature. It affects how the signature appears on paper. By analyzing line quality, we can assess the authenticity of a signature and identify any irregularities that might indicate forgery or tampering.

• Pressure: The pressure you apply while signing influences how your signature looks. It determines variations in line thickness within the signature. For instance, areas with increased pressure may appear darker or thicker. Pressure analysis is useful in verifying the integrity of a handwritten signature and detecting potential fraud attempts.

• Speed: The speed at which you sign your name contributes to the unique characteristics of your signature. Different signing speeds result in distinct visual features, such as fluidity or breaks between letters. Analyzing the speed helps establish the genuineness of a signature and assists in identifying any anomalies that may suggest unauthorized replication.

• Size and proportions: The size and proportions of your signature involve the relative dimensions of the letters and the overall layout. This encompasses the height, width, and spacing between letters. By examining these aspects, we can evaluate the consistency of your signature across different documents, aiding in the verification process.

• Flourishes: Flourishes are decorative elements that some people incorporate into their signatures, such as loops, curls, or underlines. While they add a personal touch and aesthetic appeal, they can also serve as additional features for signature verification. Analyzing the consistency and presence of flourishes helps determine the authenticity of a handwritten signature.

Verification Methods

In "Deep Learning Based Handwritten Signature Recognition" (Pokharel et al, 2020), the authors used deep learning model based on the CNN architecture to verify the signature with 95.2% accuracy.

In “Offline signature identification via HOG features and artificial neural networks" (M. Taxkiran and Z. G. Çam, 2017), the authors demonstrated 98.33% accuracy rate.

In "Offline Signature Verification Using Convolutional Neural Network" (Bonde et al, 2022) the authors used Gaussian Weighting Based Tangent Angle (GWBTA)+fine tuned convolutional neural network (CNN) and used it to classify authenticity.
#StopAIVote

I gave the examples above to show there are many ways to validate signatures, with increased accuracy, which can be deployed in Automated signature verification (ASV) systems. Which brings us back to the question: can a machine learn to perfectly fake signatures?

The Perfect Theft

Let's assume you want to rig the election using main-in ballots of either fake voters or even real voters, and make sure that no one will be able to catch you, NOT EVEN VIA A SIGNATURE EXPERT. All you have to do is the following:

1. Obtain a signature verification database.

2. Train AI/ML on the signatures.

3. Train an AI robot printing system to print a replication of signatures.

4. You own the election.

AI vs Handwriting

What I described above is FAR from being hypothetical. Let’s first look at how machine learning has been implemented to replicate handwriting:

EXAMPLE 1:

Bond, a 2015 startup in New York City created robots that can learn your handwriting and convert digital notes into personalized letters. It also worked with signatures.

A Handwritten Card, Signed and Sealed by the Latest Technology (Published 2015) The start-up Bond used robots to create personalized, handwritten notes and letters for people who want to send something special but don’t have the time. It was sold, along with its technology.

EXAMPLE 2:

The Handwriting Company was a start-up that aimed to use ML/AI to mimic handwriting by scanning the handwriting. As the BBC reporter who had his handwriting analysed by the machine and saw the way it replicated his handwriting has stated "It's not just printing the words on a paper. It's applying pressure at certain points I apply pressure".

As scribeless, the company that developed this technology states, people can’t tell the difference between a real handwriting and a machine generated one:

In an interview one of the co-founders of scribeless gave three years ago he explained:

“When we started we were like 'The only way we're gonna be able to … making sure it's efficient is if it looks realistic. If it isn’t realistic, we're wasting our time’. And so, that was opened up about a year and a half (to) two years of building out this piece of machine learning and AI, which mimics everything from ink distribution, pen pressure, ink gradients on actually how the ink sits on the page, as well as character variability, … x and y-axis variability, … the same way that every time you write ‘a’ it's fundamentally unique from the last ‘a’ you wrote, our system basically mimics that every time it processes an ‘a’ through the engine. It fundamentally looks different from that last ‘a’. So it's trying to bring all those factors in, but this isn't just a shortcut, it's actually a rather complex system which just on the output is quite efficient.”

I can hear you say “well, there might be some systems out there that were able to replicate handwriting, but signatures are a different”. This is true. For example, the start-up Bond, which allowed signature replication, used human calligraphers who provided the quality control needed for the signature to be authentic. Furthermore, in "Writer-independent signature verification; Evaluation of robotic and generative adversarial attacks" (bird et at, 2023) authors demonstrated that using a XY robotic pen plotter (iDraw 2.0 and line-us) it is possible to identify forgery. What is an XY robotic pen plotter? It’s a printer that can move right and left and apply a pen on a writing surface, such as paper (or in our case, a ballot). It lifts it up from the pen, and when it writes, it goes down and puts a standard pressure.

AS YOU SAW ABOVE, SCRIBELESS DO NOT HAVE THAT LIMITATION, WHICH MEANS THAT IT IS MUCH HARDER TO IDENTIFY IT!!! Also, the way that scribeless implements a variability to provide authenticity will make it harder for an algorithm to identify a forged signature even if you train it to identify a situation where a signature is “too good to be true”.

Brushing AI

In 2019 a Brown University undergraduate student and a staff member developed an algorithm that enables robots to reproduce human-like pen strokes just by looking at images of handwriting or sketches, with 93 percent accuracy,

In the paper, entitled "Teaching Robots To Draw", the authors shown that their system was able to replicate characters from multiple languages which it was not trained upon.

The authors also stated that "In the future, we will explore more advanced types of drawing such as drawing with a paint brush, where the speed of the stroke, the height/altitude/azimuth of the brush relative to the canvas plane all affect the drawing result. In this vein we are interested in closing the learning loop by enabling the robot to inspect what it has drawn and improve its drawing policy based on observation of what it actually drew".

Imagine using this system to train an AI system to perfect a signature replication.

X-Y AND Z

While scribeless provided an incredible realistic replication of handwriting, it does suffer one small disadvantage - the system applies the pen at a standard angle, or azimuth. Since humans don’t write in a single azimuth, which means that a signature expert most likely will be able to detect a handwriting (or a signature) based on the fact that the pen position is the same all the time (e.g. 45 degrees angle).

To solve this problem, you need a robotic printer that will be able to change the angle position of the pen, which is actually much simpler than one might imagine. In 2019, "Design of XYZ Plotter" (Kumar et al, 2019), the authors suggested a low-cost X-Y-Z plotter using Arduino as the controller. This was a DIY system that allowed a plotter with full control, including the angle.

The Perfect Signature

Based on the above, I suggest that the following model will result in an impossible to identify forgery.

1. Get a large database of offline and online signatures databases, such as the one mentioned here. Or, if you managed to steal (or gain access to) the signature database of different states in the US, use them.

2. Train AI on databases using multiple approaches.

3. Use a robotic XYZ print plotter with azimuth, pressure and speed capability to try and replicate signatures.

4. Capture the signature as a new offline signature database.

5. Use multiple signature verification algorithms to score the accuracy.

6. Based on the score, to re-train the system, input to the offline database and train it until it reaches the required accuracy.

7. Based on the above, develop a printing facility that will allow you to mass print it on envelopes, and… you own the election.

IT IS YOUR LAST ELECTIONS

As you saw, the technology to replicate handwriting is extremely advanced, and even at this stage, without any added technology, if the scribeless solution would be adjusted and adapted to signature instead of handwriting it will be extremely difficult for anyone to identify a fake signature, for sure in case of manual validation. Even if Automated Signature Verification (ASV) applications will be used, the malicious government who will be attacking the US will know the type of solution used and will make sure that the fake signatures will pass it. Since the camera used in these solutions is a “normal” camera that does not have a 3D capabilities, there is very little chance of it recognising a fake signature.

If you combine it with the Brown university AI module and a plotter that also has azimuth control, the result means that not only our elections can be stolen by any government who has that capability, but that one of the most fundamental elements of our contractual system is under risk.

The election system is the most important information system that exists in any country, because without it there can be no trust between citizens, and without trust there can be no democracy.

A malicious government can control the results of the upcoming 2024 elections, whether it will be a foreign government or the current government which is in power is irrelevant.

In fact, there is a much higher probability that the current US government will take part in such actions to assure the victory in the upcoming 2024 United States presidential election than a forien government. After all, it’s not as if the US democracy is run by those who pretend to represent the people, neither by the president, but by BlackRock. If such an operation takes place, it will cement the deep state control over the US forever, and therefore, there is a huge motivation for those who truly control the system to make it happen, especially since they have threat from political actors such as Robert Kennedy Jr., and to some extent, Donald Trump.

For such an operation to succeed requires, after the development of the AI technology, a highly coordinated effort across multiple states, and there is little chance that law enforcement forces in the US would not spot a foreign actor conducting such an operation. However, if the operation is conducted by the deep state, which is very much in bed with BlackRock and perhaps even beyond it and its kind (Vanguard, anyone?) there is very little chance it will be brought into light, especially since BlackRock controls the media.

The risk (which is the probability and magnitude of a future loss event) to the US democracy due to the development in AI and machine learning is real, is here, and UNLESS AMERICAN BAN MAIL-IN VOTING, THEIR DEMOCRACY IS OVER.

THE FINAL WARNING

IF THE USA FALLS, A NEW ERA OF HELL ON EARTH WILL ERUPT. AMERICANS - THIS IS YOUR LAST CHANCE TO SAVE YOUR DEMOCRACY. DO SOMETHING ABOUT IT. SPEAK ABOUT IT. SHARE MY ARTICLE. WRITE TO YOUR REPRESENTATIVES TO STOP MAIL-IN VOTING.

DO NOT TRUST YOUR GOVERNMENT ON THIS. IF YOU DO, IT WILL BE PROBABLY BE THE LAST ELECTION IN US HISTORY. BAN MAIL-IN BALLOTS, NOW !!!

Serving the truth with love,

Ehden