How AI can be used in forensic science: Challenges and prospects
In recent years forensic science has become a ‘sexy’ topic and area of interest. This is in part due to the hit-TV series, CSI, but also due to a proliferation of fascinating specialisations within forensics. This article introduces the contemporary landscape of forensics, and then analyses to what extent Artificial Intelligence (AI) can be of service to this industry. The challenges and prospects faced by forensic science in relation to AI-powered software solutions is discussed, touching on issues of big data, distributed environments, building statistical evidence, pattern recognition, and data sovereignty.
CONTENTS
What is forensic science?
What is AI?
What are some challenges faced by forensic sPECIALISTS?
Big data vs time
Data storage
Distributed environments
Pattern detection in differing contexts
Presentation of facts and argument validation
What are the solutions offered through leveraging AI?
Smart search
Pattern recognition
Knowledge representation
Building statistical evidence
Creating repositories
Improving communication in forensic teams
What are the potential challenges faced by AI in forensics?
Cost, storage, and hardware
Data encryption and hiding information
Balancing act: human and machine
WHAT IS FORENSIC SCIENCE?
Forensic science involves applying scientific methods and processes to collecting information about crimes and solving crimes. The first school of forensic science dates back to 1909, but the practice has been recorded as early as the 16th century with regards to pathology. In recent years, the field of forensics has exploded with highly developed specialisations, and contemporarily there are a great number of fascinating specialisations, including DNA, botany, dentistry, explosives, accounting, and cyber.
Given this diversity, forensic science draws from a many scientific tributaries, including biology, physics, chemistry, computer science, and mathematics. Forensic professionals work hand-in-glove with judicial processes. They can operate independently, with/for legal firms, or within/for the state, and they build the evidence for cases brought forward in court. Forensic science seeks to prove the existence of a crime, or who the perpetrator of a crime is, or a connection to a crime through the following key tasks:
Examination of relevant direct and indirect evidence
Administration and organisation of tests
Analysis of data
Evidence-based reporting
Accurate testimony
WHAT IS ARTIFICIAL INTELLIGENCE?
WWII British mathematician Alan Turing asked the question that has become the foundation of modern Artificial Intelligence, “Can machines think?”. By 1950 Turing had penned his influential paper, ‘Computing Machinery and Intelligence’, and as you could deduce from this title, AI is - in general terms - the endeavour to simulate human intelligence in machines. In 2004, John McCarthy proposed this definition, which has become a ‘gold standard’ for contemporary understanding of AI:
“It is the science and engineering of making intelligent machines, especially intelligent computer programs. It is related to the similar task of using computers to understand human intelligence, but AI does not have to confine itself to methods that are biologically observable.”
AI is an interdisciplinary science with many emerging approaches and sub-divisions, but one can confidently say that recent contributions in AI are influencing revolutionary changes in many industries.
WHAT ARE THE CHALLENGES FACED BY FORENSIC SPECIALISTS?
Considering the increasing number of specialisations within forensics, it is safe to say that this discussion on challenges faced by forensic scientists is not exhaustive. These are merely some pre-eminent ‘umbrella’ problems, or generalised pain points.
BIG DATA VS TIME
The first and most obvious pain point is that forensic investigations often require searching through, cross-checking, and managing huge volumes of data. This data can vary in file type, from written to legal files to payslips, to medical reports, to graphs and logs. They can also include photographs, audio, and video files. For example, a forensics team may be working on a case against a suspect involved in corporate fraud, theft and embezzlement, and this requires investigation across decades of company documents, as well as CCTV camera footage, and digital activity.
Ordinarily, this would require a small army of assistants, and the ‘search’ through each source file type (video vs emails vs accounting reports) would be a separate process. This can take years. And sometimes, by the time the research has been completed, the context of the case has changed and the relevance of the investigation at the outset has also changed. The service of justice is too often held back by the sheer volume of data that has to be managed.
2. DATA STORAGE
Linked to the point above, is the issue of storage. There is the practical issue of finding sufficient storage, but more prescient than that are the complicated considerations of data sovereignty and security. Naturally, the evidence brought forward in a criminal case is sensitive and highly fragile. If it gets into the wrong hands, the implications can be really damaging; it could be altered/tampered with, leaked to the wrong sources at the wrong time, or deleted/wiped altogether.
Moreover, there may be legal requirements brought forward that dictate that the data cannot leave the country in which the investigation is being carried out. This could mean that cloud storage solutions are out of the question, or have to be customised. Where and how the data for a case is stored is a complex technical question. Custom storage is also very expensive.
3. DISTRIBUTED ENVIRONMENTS
This is a pain point for digital forensics in particular. There are two possible aspects to this. The first is when investigating a cyber crime it may be discovered that the source(s) of the crime are not traceable to one environment. Meaning, that the ‘cyber attack’ came from different sources, different end points, in different countries. This makes it very hard to trace.
The second example is taking into account that many forensics investigations these days require at least some digital tracing. A criminal or criminal organisation leaves traces not just through fingerprints and dentistry, but through their activities online. This fast developing specialisation of forensics requires tracing and tracking online activity, and if the activity is dispersed across many environments and end points, it is particularly challenging, time consuming and requires the cooperation of many platforms and software companies that may treat user data in different ways.
4. PATTERN DETECTION IN DIFFERING CONTEXTS
Scientific disciplines all have a certain set of rules and logic. The assumption often made from within the discipline is that the subject matter of study follows the same epistemological framework. But this is not the case. There are numerous systems of knowledge; there are many forms of logic. When investigating a case, forensic scientists must search for patterns and coherent links across a vast sea of possible data. Oftentimes this data is collected from different disciplines (DNA/ botany/ dentistry/ art), all with their own assumptions and knowledge frameworks.
How do you then coherently and accurately link a pattern across this diverse landscape? Which epistemological framework is the meta-framework, to unify and collate them all? Does an assumption from one discipline alter the way findings look from another discipline? Finally, when all these layers of patterns and knowledge within science are projected onto society and human interaction, are they still relevant, integral and valuable insights, or does the irrationality and unpredictability of human decision-making rub up against the science?
5. PRESENTATION OF FACTS AND ARGUMENT VALIDATION
Once the evidence has been collected, findings need to be collated into reports and translated into understandable and useful narrative. This narrative becomes part of the argument served to a judge or jury. The journey from scientific data collected through a specialist area of forensics to a digestible narrative is not straightforward and takes a lot of skill and time.
A good example is pathology and medical forensics. How do the small details of anomalies or injuries picked up in an autopsy relate to the narrative woven into an argument? Or, with regards to accounting forensics - how does numerical data synthesise with argument validation? Reporting and argumentation is another step weighed down by sheer volume and complexity on the journey a forensic investigation takes.
WHAT ARE THE SOLUTIONS OFFERED THROUGH LEVERAGING AI?
There are a number of ways that Artificial Intelligence can assist in forensic investigations and reporting. Some examples are:
Smart search
A method of AI, machine learning (ML), is particularly useful in data analysis. Think ‘smart search engine’. Using natural-language processing, ML powered search engines are able to survey enormous data sets, with varying types of files and information, and bring forth semantic search results. This means that they not only snag on and pick up exact keywords or search terms, but even likened (semantic) and linked data. This can cut down the lifecycle of a forensic investigation significantly, sharpening months or years of investigations down to mere weeks.
2. Pattern recognition
This identifies certain clusters of data in an investigation. ML-powered pattern recognition tools can unearth a logic of similarity across a set of data. This can be over images (using facial recognition, for example), audio and video, and other document file types. This can help forensics to link findings, or data across findings more quickly and - potentially - more accurately.
3. Knowledge representation
Knowledge representation is a field within AI that is concerned with how knowledge works within a domain. Using machines, all knowledge within that domain is collected and mapped to establish how concepts relate to each other, thus defining what the rules are that control how they behave. This is particularly useful in cross-disciplinary forensic investigations, and in digital forensics when investigating cyber crime. Mapping knowledge systems can more accurately inform how and in what ways is is appropriate to link evidence across disciplines, and to work out cross-referencing relevance. In digital forensics, it helps detect the origin of an action or attack.
4. Building statistical evidence
AI tools for building analytics and reporting templates are vastly improving efficiency in all professional fields that require a lot of data to be translated and used to make inferences and decisions. AI-powered analytics are updated in real-time and can automatically render insights on data throughout an investigation. This also makes it easier for those involved in the judicial process to digest forensic reports and findings.
5. Creating repositories
Storing sources in a way that is coherent and quick to navigate is not easy when the data set is so huge. However, auto back-linking and data management tools powered by AI are transforming this area from a headache to a breeze. Using AI, all sources can be automatically linked to origins and other relevant files or sources. Moreover, AI assists in coherently grouping data sets for storage so they are more intuitively recalled and rediscovered when the time is right.
6. Improving communication in forensic teams
Over the course of an investigation, the team of forensic experts involved in the case can be numerous and changeable. Keeping track of everyone’s contributions and organising communication across the team(s) in an efficient way is very important. It is also important to provide the correct type of interface for team members who are in different disciplines to be able to comment on and contribute to others’ work. Workflow and document annotation tools with ML and AI injection are the answer. Team organisation and management has never been this easy, and many of these interfaces link directly to the cloud (custom or public) for storage of information so team members can access it from anywhere.
WHAT ARE THE POTENTIAL PROBLEMS FACED BY AI USE IN FORENSICS?
Cost, storage, and hardware
The hardware needed to crunch and store big data, and the software skills needed to package and maintain these services are usually very costly. It is worth noting, however, that in terms of software, there is a very exciting African AI development curve, and this has enabled some excellent, more cost effective AI software solutions (notably, Doc Insights). Nevertheless storage and security issues remain unresolved when it comes to AI in forensics. If a case is particularly sensitive, data sovereignty issues can entail difficulties in storing data and even using AI tools that are hosted in countries other than that which the investigation is being carried out in.
2. Data encryption and hiding information
This problem arises particularly in cyber crimes and digital forensics. As much as AI tools for helping in forensic investigations develop, other technologies that assist in hiding information, or creating digital walls beyond which investigations cannot penetrate. And so there is a race between the development of solutions for investigation, and solutions for contrary intent. It is likely that the two have a parasitic relationship and that this problem is not solvable, but inherent to the field itself.
3. Balancing act: human and machine
When all the fascinating and cool forensic investigation is done, the end game is legal. Judicial processes are fraught with ethical conundrums and problematics - world over. The service AI provides for the judicial process has to mesh with the rules of the game of law and human ethics. AI cannot (at this stage anyway) replace the human endeavour to provide justice and equality. AI cannot replace the complex and imperfect art of justice, and so checks and balances need to be put into place to ensure that AI powered findings and contributions are handled appropriately. The governing of and legislating for AI contributions to forensics will be an ongoing dance in years to come.
For more information about how DocInsights can assist forensic scientists and forensic investigations, get in touch with the Doc Insights team and book a demo. Doc Insights offers subscription solutions and custom enterprise solutions.