Oracle has acknowledged a still-unpatched database vulnerability after proof-of-concept code was published.
The company advised users to apply workaround measures and configuration changes.
The serious vulnerability earned a 7.5 CVSS score and was remotely exploitable without the need for authentication.
It affected 10g and the most current version 11g of Oracle Database.
Researcher Joxean Koret reported the man-in-the-middle flaw to bug-bounty program iSight Partners four years ago, which shared details with Oracle per its reward program specifications.
He disclosed the proof-of-concept code after he thought Oracle had fixed the flaw in a patch issued two weeks ago and was given credit in the company's Security-In-Depth program.
But in an email exchange he discovered the defect was only repaired in future versions of the database.
He said attackers can exploit the flaw to "sniff any connection" made to the database without the need for credentials, and can also inject malicious commands.
He added that he is not aware of any in-the-wild attacks underway.
Oracle tried to fix vulnerabilities first in "the main code line" and then "backported" through its quarterly security updates.
But sometimes backporting is not possible due to heavy amounts of code required or because doing so will cause serious performance issues for customers.