Automated Forensic Examination: Issues and Challenges

Expert Profile

Foreground

The session has been started with a brief and insightful introduction to the expert Dr. Rajesh Kumar sir. The session was then handed over to Dr. Rajesh sir for sharing his talk on the title automated forensic Examination, Issues, and Challenges.

He open his talk by wishing Gandhi Jayanti as the session was held on 2nd October 2021. Talking about the automated forensic examination Dr. Kumar remembered a story from one of his expert talk held in 2009 at IIT Delhi, there a student asked that if there are different automated techniques are being used so why still most of the forensic laboratories used manual methods for all the forensic examination?? It has been many years since then but still, the scenario has not been changed much. So in this session, Dr. Rajesh Kumar sir have discussed in this session about automated forensic examination issues and challenges

So here are some key points discussed in his talk

Driving Force for Automated Methods

If we see the history of forensic examination in India and outside it's now more than 100 years. We are doing lots of things we are talking about automation but then say somehow we are lagging so there must be some reasons behind and there are various report which criticized the forensic examination many times, not all but few of them are criticized much time

So let’s discuss a very recent report PCAST Report (2016),  which was regarding forensic examination that whether it is going accurately or what is the problem and challenges facing out by forensic examiners in the forensic examination process. The NRC Report in 2009 was another one in which basically a recommendation of problems in forensic examination in the United States was detailed and reported. But the most significant report, we can say which changed the forensic world was the Daubert Report in 1993. Before this report, there was a case of polygraph when it came across the court of law in the united stated in which certain boundaries have been set to know that what kind of scientific evidence should be acceptable. So it was all about general acceptability but Doubert has changed the entire forensic world pertaining that if you want to give any scientific evidence report in a court of  law, they should go with certain procedure and should satisfy certain criteria and for that and the court must have some gatekeepers to see that whatever scientific evidence is being given they all must be fully proofed. Still, there are few things which are observed in most of the reports in common are

• There is a lack of proper scientific research in forensic setup
• Lack of objectivity in analysis and opinion
• Lack of statistical basis
• Bias in manual examination

So these all were the problems which were highlighted in all the report. As per Daubert, there are 5 criteria that must be fulfilled for whatever we are doing or whatever scientific evidence we are presenting to the court of law, that should be followed

1. It should be testable
2. It should be peer-reviewed
3. There should be a Standard Operating Procedure which should be followed
4. The potential error rate should be minimized and validated
5. There should be general acceptability among the community and standardization about the technique which are being used for scientific examination

What can be automated?

Now after this we need to know that what forensic examination or a part of the forensic examination can be automated

• Recognition/Linking of Impression evidence
• Recognition/Linking of trace evidence
• Means of person identification
• Questioned document
• Analysis of digital evidence

So almost everything where the decision has to be made are needing to have automation in the analysis process to bring the error rate around zero

Automatic process

Now let’s know what kind of automatic process is there. So basically when we talk about automation we are talking about

• Classification/clustering
• Regression: it is about prediction,

Data Acquisition and pre-processing

Data acquisition and pre-processing is to make the data into a suitable form. This can be done in the following steps

1. Feature Extraction: Feature extraction is basically when we want to represent anything in terms of some characters for example let’s say we have a banana and we want to represent it as an individual identity,  so here we may look for its shape or maybe the color and the texture So basically the feature extraction is not a very complicated task. The only thing we need to is to represent the object well.

2. Feature selection: Now that we have extracted the feature but whether these features are good or bad for example let’s say that work has been given to a group, 5 people are doing some work and others are doing other work and both of them are taking 5 hours because they have the same work

There are 3 types of technique

• Filter: Filter means we are section each feature individually like if we have 10 features then we will evaluate each one of them separately and there should be no interaction between any feature
• Wrapper: Here we take care of the interaction between the features which means when we are evaluating 2 features, we have to see that which of them are good for further analysis and which can be discarded from the analysis.
• Embedded:

3. Classification:

Classification is like when we have to classify two different entities on basis of their class feature for example if we have to identify banana and apple. Here we need to learn the boundaries from which we classify them. We can learn to classify with using any of the following methods such as

• Unsupervised Learning: unsupervised learning is basically when we learn t differentiate two things based on their common similarities and dissimilarity in their shape, size color, and other features without any in-depth knowledge about that thing. Here the learning is not supervised or supported by anyone else’s knowledge.
• Supervised learning: supervised learning is when we learn with someone’s continuous supervision for a longer duration of time. Suppose if we again take an apple, for example, we have learned to identify it as an apple because we have been shown and told many times that this fruit of particular shape, size, color, and morphology is called an apple. So now whenever we see a fruit similar to that particular feature, we can identify it as an apple
• Reinforcement Learning: In this learning as the name suggest learning occurs with trial and error with punishment and rewards

4. Performance evaluation:

For incorporating the automated techniques in any case for forensic analysis, there are a few thing that should be taken care of to evaluate the performance and accuracy of the instrument

• False Positive
• False Negative
• Equal error rate
• Precision
• Recall
• F score

Dr. Kumar has also discussed some really interesting automated methods which he and his team have collectively designed for research purposes only. He has discussed the usefulness of those automated methods such as in Fraud detection in check forgeries, Handwriting analyses, Fingerprint comparison, and many more. To know the importance and significance of automated methods in forensic examination in detail, I suggest visiting the YouTube channel and hearing our Dr. Kumar itself.

Challenges:

All the instruments used by Dr. Rajesh Kumar and his team are for educational and research purposes and are really beneficial to save time, assure greater accuracy, and have fewer errors in examination results as compared to the manual methods. But again every coin has two sides. If automation has innumerable benefits in forensic examination then in the other side of the coin this method also has some challenges and limitations which are faced by the forensic examiners. Some of those challenges are motioned below

• Storage and protection of the database
• The black-box nature of laboratory personals which means that in the laboratory, not every person s aware and of these automatons and working with automated techniques is like out of the comfort zone for them and this attitude makes the automated forensic examination as a big challenge in most of the forensic laboratories.
• High Cost- again the next and foremost challenge is the higher purchasing and maintenance cost of the automated techniques which is not affordable for most of the forensic science laboratories
• The next challenge is the dependence to vendor due to unawareness of the technicality of the instruments so f the machine shoes any issue in between or need an updating the laboratories are completely dependent on the vendor to solve the issue and get pupation.

Path Forward

According to Dr Rajesh Kumar, for a futuristic approach in forensic examination, the examiner must appreciate the interactive automated system to incorporate in the analysis part for forensic examination of different cases. There are few suggestive measures which if applied with caution then the automation and automated techniques can become a saviour in several cases.

• There must have a proper storage and protection facility for the database
• There must have a Standard Operation Procedure for all the processes which are being followed
• And there must have quality control assurance for all the automated methods used in the laboratory and they all must be updated from time to time.

Conclusion

The session was such an amazing one. And it was concluded with a quote by Dr. Kumar that “Automation is not the replacement of expert”, these techniques can help in speedy disposal of cases but still there is always the need for an expert for all the observation and a second opinion must always be taken for all the analysis.

Personal Note

For me, this session was so interesting and entailed numerable aspects which should be taken into consideration while incorporating automation in forensic analysis. I suggest all my readers to go through the session recording at YouTube channel Forensic365 to get a deeper insight of this session.