Convenience and security don’t always go hand in hand. For example, over the past several years, citizens have urged their city, state, and county governments to invest in e-government initiatives so basic government forms, applications, services, and official records can be accessed over the Internet. While these initiatives offer citizens a convenient outlet for information and allow government entities to serve more of their constituents with the same number of personnel, they can come with a hefty price. Many records that are public in nature (i.e. deeds, marriage licenses, death certificates) can contain private information such as Social Security numbers, credit and debit card numbers, and bank account numbers. Making this type of information public is an invitation for identity theft.

This threat caused the state of Florida to issue a redaction mandate. Under Florida’s current public records law, citizens are responsible for requesting that private information be removed (or redacted) from public documents. As of Jan. 1, 2007, the responsibility will be shifted to the clerks of court to ensure that all sensitive information is kept confidential under Florida’s open-record laws.

MANUAL DATA REDACTION
TAKES RECORDS OFFLINE
This mandate posed a problem for the Marion County Clerk of the Circuit Court. The county already had more than 7 million official records dating back to 1965 electronically stored in an image archive solution developed by NewVision Systems Corporation. These images were also already available to the public via the Marion County Clerk’s Web site when the redaction mandate was issued and were frequently accessed by title companies, insurance companies, and insurance offices. Redacting sensitive information from these 7 million documents manually was not a feasible option.

“We wouldn’t have been able to redact this information manually and still provide online access to the records,” says Jack Suess, chief deputy clerk of administrative services for the Marion County Clerk of the Circuit Court. “We would have had to remove all the records from the Internet and revert back to the old days when people were actually required to come into the courthouse to ask for a particular record. Physically redacting sensitive information when a hard copy request was made was the only way we envisioned this process could be accomplished manually, and this would not have sat well with our staff or our citizens.”

The Marion County Clerk immediately began investigating technology alternatives. “We started the discovery process early because no company we knew of offered a technology that addressed this type of redaction requirement,” says Suess. “Plus, we didn’t want to outsource the redaction process because of the sensitive information involved. We wanted to maintain control of the project to ensure the information didn’t end up where it wasn’t intended.”

BE A GUINEA PIG FOR
UNSTRUCTURED DATA RECOGNITION
The Marion County Clerk’s office decided on a solution developed by systems integrator Computing System Innovations (CSI) called IntelliDact. “Other vendors claimed they had software that would identify sensitive information and automatically redact it from all our records,” says Suess. “However, when they performed a live demonstration, they would use the same document type over and over again. This seemed staged to us. CSI, on the other hand, used our actual records in its demonstration and proved its software would work, regardless of document type.”

CSI worked closely with the Marion County Clerk’s office to customize IntelliDact to address its specific business rules. “CSI had the technology components down, but needed to become familiar with the terminology of our industry, the different document types we process, and the private information we need to redact,” adds Suess. “For example, it needed to know if a federal ID number was the same as a Social Security number to determine whether or not it should be slated for redaction.”

Beta tests of the software were then run separately on each of the distinct document types the Marion County Clerk processes. A sample of 500 documents of each type was run through IntelliDact to ensure the redaction process was successful. With the CSI solution, information is extracted from document images using Kofax Ascent Capture and delivered to IntelliDact. IntelliDact then uses unstructured data recognition technology to scan an image for sensitive information and automatically slate it for redaction. This process is based on business rules provided by the Marion County Clerk’s office. For example, the business rules may prompt the software to search for keywords such as Social Security Number, SS#, bank account, and credit card number and then look to the left, right, above, or below to locate that number and mark it for redaction. The technology can also find free-form Social Security numbers without proximity to a keyword based on number patterns. For example, the common Social Security number pattern is three digits-two digits-four digits. IntelliDact also comes equipped with an ICR (intelligent character recognition) engine so it can identify handprinted numbers in addition to machine-generated print.

Once IntelliDact identifies a document contains information that may need to be redacted, it places the image in one of three queues. A red queue indicates highly questionable information that requires manual sight verification by a Marion County Clerk employee before completing the redaction process. A yellow queue contains documents the system suggests a verifier review before completing the redaction process. Finally, a green queue contains documents that the system is confident it has redacted properly. It is usually unnecessary for a verifier to review documents in the green queue.

PROCESS 7 MILLION RECORDS IN
SEVEN WEEKS WITH AUTOMATED REDACTION
Once all beta tests of IntelliDact were complete, CSI copied the Marion County Clerk’s entire database onto independent servers and began the redaction process on all 7 million records in its back file of images. “Even though our statistics indicated that only 8% to 10% of our records would actually contain information that would require redaction, we had no way of knowing which documents they would be,” says David Ellspermann, clerk of the circuit court for Marion County. “Therefore, we had to send our entire database through the redaction process. However, we know that certain document types, such as death certificates, are likely to contain sensitive information, so we streamlined the process by using the recognition technology in IntelliDact to classify these document types and process them up front.”

To ensure it would comply with the Florida mandate on time and could accommodate any unforeseen processing delays, the clerk’s office started the redaction process with records from the most recent year and worked its way back to the older records. However, the Marion County Clerk of Circuit Courts didn’t have to worry about any technical hitches. “Using multiple servers and unused PCs to run the unstructured data recognition process, we were able to redact information from our entire back file of 7 million records in six weeks and four days,” says Ellspermann. “This made us the first county in Florida to successfully comply with the redaction mandate.”

Online records that contained sensitive information have since been replaced by newly redacted versions. Furthermore, the records didn’t need to be removed from the site for long periods of time in order to make this transition. The Marion County Clerk still retains the original, unredacted images for individuals who request certified copies of their personal records.

New records processed by the Marion County Clerk currently go through the same process as the 7 million back file images. Documents are scanned, recorded in the NewVision imaging repository, and then sent through the redaction process before they are posted to the Web. However, CSI and NewVision are currently integrating their technologies to the point where records can be redacted as they are scanned. This capability will further streamline the process and allow redacted records to be posted on the Marion County Clerk’s Web site in a more timely manner.