Anomalous traffic volume can be used for identifying network threats and faults. Denial of service (DoS) and quality of service (QoS) are two contrasting problems of anomalous network traffic. DoS exploits malicious traffic to hinder service availability to normal users, whereas QoS determines if the service provision quality has reached the preset agreement. This paper proposes a unified ant agent framework for identifying the source of these problems: IP traceback for DoS attacks and fault localization for QoS violations. Numerous studies have investigated IP traceback techniques for identifying spoofed IP addresses of attackers. These techniques can identify the attack path from the victim to the attacker. Metaheuristic algorithms that consider slight increments in traffic volume (SITV) are rarely studied for solving the IP traceback problem of DoS attacks. We investigated the malicious and nonmalicious situations for the QoS attack and QoS fault localization problem. This paper proposes a novel ant colony optimization (ACO) method for fast filtering, DoS threat source identification, and QoS fault localization (unified threat identification and fault localization by using ACO, UTFACO). The UTFACO framework was compared with the probabilistic packet marking approach and conventional ant system algorithms. We com- pared the efficiency of UTFACO with and without a bloom filter (BF). The framework was verified in the QoS attack and QoS fault experiment environments. This study showed that attack or fault detection and identification procedures can be designed and implemented practically. The tests used the dataset of the network topology from the DARPA repository with two cases: one is a general experiment, and the other has various levels of SITV. Perfect accuracy can be achieved for the general experiment, and more than 90% accuracy can be obtained for various levels of SITV. The datasets of the QoS attack and QoS fault were obtained from a real network. Precise fault localization is achieved due to the high detection rate obtained. The results show that UTFACO is an efficient and accurate framework. Moreover, the computation time is considerably reduced by using UTFACO with the BF, and the time is less than five seconds in the framework. Our proposed frame- work is robust and can solve the problem of identifying the IP address of an attacker and detecting the fault location.