Regulator Trai has directed all mobile operators to share data derived from their artificial intelligence (AI)-based anti-spam solutions within hours on the common blockchain-based platform and act against spammers, even in the absence of any complaints.
The directions by the Telecom Regulatory Authority of India (Trai) were issued on February 27, and firms must comply within 30 days.
While the regulator hasn't specifically asked for blocking numbers of potential spammers after telcos raised concerns on the draft rules, the regulator has fixed accountability for both the originating and terminating telco-from where the call or message originates and terminates-to coordinate among themselves and initiate action.
As per the directions, every terminating mobile firm, through its AI/ML (machine learning)-based spam detection system, should identify and flag the calling line identification (CLI) or mobile number of the sender as "suspected spam CLI" based upon the behavioural parameters as specified in the system. Further, immediately upon such flagging and in any case within two hours of such flagging, the terminating operator should share, through the distributed ledger technology (DLT) or blockchain platform, the flagged CLI with the concerned originating operator.
The originating operator is then required to reach out to the sender of the CLI informing him about flagging of the number as suspected spam and also the need to ascertain know your customer (KYC) identifiers of the sender of the CLI. The data should be shared on the DLT platform so that all telcos can identify the telecom resources allotted to the sender and check if other mobile numbers of the same sender are also identified as potential spam by their respective systems.
In case it is ascertained that five or more CLIs of the sender have been identified as potential spam within 10 days, action should be initiated against the sender.
Telcos had earlier cautioned against using AI-based data to act against spam as the data isn't entirely accurate and may tag or flag genuine numbers. The telcos want to continue with the existing mechanism when action is based on complaints by users.
Further, AI solutions by telcos are not similar and have different technical parameters, so sharing of data can have technical limitations.
Trai, however, said, "nothing in this direction shall require disclosure of proprietary algorithms,
source code, model architecture or internal risk-scoring methodologies of the AI-based UCC (unsolicited commercial communication) or spam detection system deployed by any Access Provider."
The authority feels that while the telcos have implemented the spam solutions, just alerting the subscriber without backend enforcement does not act as a deterrent. "...enforcement has, thus, remained predominantly complaint driven," it said. Further, Trai said around 85% of spam complaints are reported against unregistered telemarketers, and therefore, effective containment of such spam senders requires a calibrated leveraging of AI/ML-based network intelligence deployed by telcos.
The directions by the Telecom Regulatory Authority of India (Trai) were issued on February 27, and firms must comply within 30 days.
While the regulator hasn't specifically asked for blocking numbers of potential spammers after telcos raised concerns on the draft rules, the regulator has fixed accountability for both the originating and terminating telco-from where the call or message originates and terminates-to coordinate among themselves and initiate action.
As per the directions, every terminating mobile firm, through its AI/ML (machine learning)-based spam detection system, should identify and flag the calling line identification (CLI) or mobile number of the sender as "suspected spam CLI" based upon the behavioural parameters as specified in the system. Further, immediately upon such flagging and in any case within two hours of such flagging, the terminating operator should share, through the distributed ledger technology (DLT) or blockchain platform, the flagged CLI with the concerned originating operator.
The originating operator is then required to reach out to the sender of the CLI informing him about flagging of the number as suspected spam and also the need to ascertain know your customer (KYC) identifiers of the sender of the CLI. The data should be shared on the DLT platform so that all telcos can identify the telecom resources allotted to the sender and check if other mobile numbers of the same sender are also identified as potential spam by their respective systems.
In case it is ascertained that five or more CLIs of the sender have been identified as potential spam within 10 days, action should be initiated against the sender.
Telcos had earlier cautioned against using AI-based data to act against spam as the data isn't entirely accurate and may tag or flag genuine numbers. The telcos want to continue with the existing mechanism when action is based on complaints by users.
Further, AI solutions by telcos are not similar and have different technical parameters, so sharing of data can have technical limitations.
Trai, however, said, "nothing in this direction shall require disclosure of proprietary algorithms,
source code, model architecture or internal risk-scoring methodologies of the AI-based UCC (unsolicited commercial communication) or spam detection system deployed by any Access Provider."
The authority feels that while the telcos have implemented the spam solutions, just alerting the subscriber without backend enforcement does not act as a deterrent. "...enforcement has, thus, remained predominantly complaint driven," it said. Further, Trai said around 85% of spam complaints are reported against unregistered telemarketers, and therefore, effective containment of such spam senders requires a calibrated leveraging of AI/ML-based network intelligence deployed by telcos.
